I would sugest the the auto-ionization mechanism the way you describe it is only one possible way of quenching particularly relevant for combinations of hosts and activators where the activator has some excited state very close to the edge of the conduction band. Only in this, not too often encountered, situation the probability if thermal ionization can be higher than, for example, usually considered quenching by multi-phonon emission. But, let's assume we have this rare situation. Electron promoted to the conduction band is delocalized and can move inside the crystal until it  either falls directly in the valence band (and emits light at much shorter wavelength than from the activator ions) or, very probably, it will find another recombination cite. If this recombination level is again another ion of the activator, the story starts from the beginning - it either goes to the ground state and emits light as we want it to, or otherwise there is again a good chance that it "ionizes" back to the conduction band and continues its journey across the crystal. However, this new recombination level can also be a defect providing non-radiative relaxation route. So, auto-ionization at least to some extent flushes the electrons "down the sink" - by giving them the chance to find a quenching defect, or, with probably not too high likelihood, also gives them an opportunity to emit light with a "wrong" wavelength. That means for you, that auto-ionization in any case does not improve the efficiency, and, as long as there are no crystals without defects, it quenches the luminescence. The higher the probability of auto-ionization the stronger the quenching.

I would add also another possibility: defects can act as traps for electron, capturing them for sufficiently long times to be considered 'lost' from the luminescence point of view. Only for sufficiently high temperatures, the electron can be released back into the CB: this give rise to the thermoluminescence phenomenon.

Roman, I'm not really sure that rare earth excited states close (or in) the CB is really that uncommon, though I agree with you that multiphonon relaxation is likely more effective.

Alright, I have to admit that I worked only with certain types of wide bandgap hosts and could not recall much on auto-ionization from the books I read. Would be nice if any one of you could recommend me a good literature on auto-ionization or maybe some examples of such materials.

Dear Roman,

Thanks for the suggestion.

Indeed, there are two models existed to explain the thermal quenching of LED phosphor, the 5d-4f crosspoint model and auto-ionization model. The quenching mechanism of the multi-phonon emission you mentioned should be related to the former one. From my perspective, I don't think this model is the reason for the thermal quenching of phosphor, (especially for the Ce3+ and Eu2+ doped. since the difference of the vacuum refereed binding energy for the 4f and 5d is large.)

For the auto-ionization model, I agree with your opinion. However, if we compare the luminescence mechanisms of phosphor and scintillator, we can find the difference. For scintillator (also Ce3+ and Eu2+ doped), the electron in CBM is captured by the 5d state of the rare earth ions, which combines with the 4f state (has captured a hole from VBM), giving a emission. If we follow this mechanism, in the case of phosphor, we can deduce that the promoted electron in CBM has a great possibility to recaptured by the 5d  state of rare earth ion. Therefore, I don't think it is reasonable to say when the 5d electron of rare earth ion is promoted into CBM, the direct quenching will occur. Anyway, I agree that the electron in CBM is much more delocalized than  the electron in 5d state.

Best wishes,

Yongchao

Dear Federico,

Thanks for providing the information about the thermoluminescence. Such phenomenon is related to the mechanism of persistent phosphors, which is usual obtained through the co-doping rare earth ions, like Eu2+, Dy3+. So, you also believe such idea can be used to explain the thermal quenching of phosphors for white LEDs?

Best wishes,

Yongchao

Dear Yongchao Jia,

I saw your question today and would like to introduce a contrarian viewpoint. In my opinion your introductory paragraph contains a category error:

Speaking about the rare-earth activated phosphor, we can think that we introduce some localized state (4f and 5d) state in the energy scheme of semiconductor.

The multi-electron terms that describe atomic energy levels are different in nature from the single-electron energy scheme (band structure) of a semiconductor. States in the gap (donors, acceptors, traps) are also single-electron states, spread over a range of k-values, since they are localised. Equating such a state with an atomic term is at best a convenient fiction.

Your question relates to temperature quenching of luminescence and you may be interested to read a short commentary I wrote to mark the 25th anniversary of Favennec's seminal work on this topic, as attached.

Regards,

Kevin

Roman,

there are a series of articles written by Dorembos and his group which report the position of 4f, and partly 5d, rare earth levels in various insulating matrices with energy gap > 5, which might give you a nice overlook to the question. In particular is is shown that for several of these ions both 4f and 5d levels can be deeply inside the CB. A few examples of  ionization under illumination at different temperatures can be easily found for LSO:Ce (lutetium oxyorthosilicate, and similar compositions with Y) and YAG:Ce (YAl garnets). At the moment I cannot give you references, but you should not have problems in finding them.

Yongchao,

The presence of trap can give a contribution, but surely isn't the main responsible for luminescence thermal quenching. Moreover, you have to consider that the stored charges will be reemitted as soon as the sample temperature is increased: this can appear as peaks in the luminescence efficiency as a function of temperature, though I admit that this is not something that clearly occurs in photoluminescence.

I tend to disagree with your comment about the multi-phonon relaxation in Ce3+ and Eu2+: the vacuum referred binding energy is not a good measurement for whar we are discussing. You have to consider the (morse like) electronic potentials for the 4f and 5d levels, their likely crossing. This substantially lowers the energy difference. You can have an idea by looking in the classic book by Blasse and Grabmeier (Luminescence of solid), in which the multiphonon relaxation is very well described. You can also have a look to Garcia Solé' An introduction to the optical spectroscopy of inorganic solids

Cheers

Dear Prof. Kevin,

Thanks for the comment.

I have some personal problems to understand your comment, but not directly link the topic discussed here. I have sent you an email. At present, I didn't change the introduction part.

Thanks for the nice paper you mentioned.

Best wishes,

Yongchao Jia

Dear Federico,

Thanks for the comment. Indeed, you are correct. The difference of vacuum referred binding energy of 4f and 5d state is not suitable for the exclusion of 4f-5d crossing model, but the difference of host-referred binding energy of 4f and 5d state.

The 4f-5d crossing model is very popular model to explain the thermal quenching of phosphors. However, this model sometimes doesn't work, for example, to explain the thermal quenching of Eu2+ ion in Ba3Si6N9O4 and Ba3Si6N12O2. Please find the reference: doi:10.1016/j.optmat.2012.11.014.

Why the model of multi-phonon emission is much effective to explain the thermal quenching of phosphor? Are there some other references to compare the model of 4f-5d crossing and auto-ionization?

Best wishes,

The question is vey complex, 1) you have to obtain the activation energy of quenching, this is very important and 2) to think for an explanation. 

Dear Yongchao,

The 4f-5d crossing model is highly effective in explaining the thermal quenching for most of the luminescence ions. Eu2+ in this respect is a rather odd one. But you also have to consider the relative position of the Eu2+ ground and excited states with respect to the delocalized bands, which for this ion tend to be closer to the CB than for others (see Dorenbos model). I'm not aware, though, of comparisons betwen the two models, though I'm firmly convinced that both of them might occurr at the same time: it is really a matter of finding which contribution prevails.

Regards,

Federico

Dear all,

Thank you very much for the comments.

Beside we have mentioned the two quenching mechanisms for the luminescence phosphors. I think we also should consider  another process,  5d-CBM crossing model. It means that the structure geometry can be changed (absorb phonon), which results in the system reaching the crossing point of 5d state and CBM.

What do you think?

Also, are there  some ways that we can measure or theoretically calculate the energy barrier of the 5d-CBM  (4f-5d) crossing model?

Best wishes,

Yongchao 

Yongchao,

" [...] 5d-CBM crossing model." I frankly don't see the difference with the autoionization model, since it does involve the presence of phonons to 'promote' the electron in the conduction band.

Dear Federico,

Indeed, the model should be similar with the auto-ionization process. However, the structure geometry for the two models seems to different.

What is your opinion?

Best,

Yongchao

Dear Yongchao,

In which way the geometry for the two model seems different?

Regards

Dear Federico,

Here is our recent paper, which may be helpful.

Best wishes,