1 Background Literature and Equipment

The current issue of this journal considers the many achievements of Professor Watanabe. These are considerable and the companion articles provide detailed examples. In an overview of a long scientific career, there is an opportunity not merely to consider his many achievements, but also to place in perspective the concepts, equipment, and facilities of anyone starting a career at that time. He has a considerable reputation both in nuclear sciences and areas related to luminescence and archaeology. Selections of his many publications are listed in various sites [e.g. 1, 2]. Current fashion is to count citations as a monitor of achievement but this is unrealistic for earlier generations, as the scientific community was small and strongly location dependent. It also overlooks the true importance of publications which in many cases appear to have limited citations, but on reference to journal details of references [1, 2], these sometimes cite hundreds of downloads. This is a true measure of the impact of the work.

The Watanabe luminescence thesis work (PhD 1961) was made during the end of the 1950s, and it is extremely difficult for a modern young student (or indeed most of their supervisors) to appreciate how our science, which addresses topics such as luminescence and imperfections in insulating materials, has changed within his long lifetime. The following comments therefore benefit from the contrast of experience from those who were postgraduate students of these topics in very different eras, from the 1950s (Townsend, PhD 1961) and early twenty-first century (Wang PhD 2007). Note that early texts on this subject [3,4,5,6,7,8,9,10] all postdate the earlier theses, a situation that is never encountered by modern students of luminescence.

A very positive aspect of early research in solid state physics was that before about 1960, all Physics Abstracts were in a single volume each year. When searching for a particular reference abstract, one inevitably saw many unrelated articles, and stopped to read them. This offered not just inspiration, but it also gave an extremely broad perspective of physics. The modern approach, with a computer search, merely finds a chosen article, or a narrow list of highly cited articles. Outliers with new ideas, or few citations, may be hidden many search pages later, and unrelated abstracts will never be seen. Such literature search digressions have often stimulated a wide range of ideas in later years and a confidence to diversify. For Watanabe, this is apparent from the spread of his luminescence studies over materials from synthetic crystals, to natural minerals, and archaeology (in addition to his nuclear physics contributions).

Modern students will find it extremely difficult to comprehend the consequences that, without semiconductor electronics, there were no calculators, computers, electric typewriters, photocopiers, web sites, etc. Further, since the 1950s, it was just postwar, funding was minimal, and equipment was frequently designed by the student (or supervisor) and constructed in the laboratory workshops. The positive side of needing to design and build equipment was the confidence and ability to devise novel equipment and attempt non-standard experiments. By contrast, later postgraduates invariably use commercial equipment, but definitely benefit from the ability to use computer data acquisition and signal processing. Nevertheless, there can be scientific weaknesses in commercial software packages, and if they are widely used, only the very best scientists will recognise them, especially if the same faults are included in the data of other authors.

In the 1950s, data, such as optical absorption, were collected wavelength by wavelength with a monochromator, by setting a zero, then a 100% level, inserting a sample into the beam, and manually recording transmission values. Subsequent processing to draw graphs etc. was also by hand. Luminescence studies started to use photomultiplier tubes, but their performance was far below that of current detectors (especially at long wavelengths). However, home-constructed equipment could be successful. For example, Fig. 1 contrasts part of the spectra from a ruby (Al2O3:Cr) excited either with electrons or light. Since these were weak long wavelength signals, the detection was via infrared-sensitive photographic film. Spectra were dispersed in a laboratory-made Rowland spectrometer using a 1-m radius curved grating [11]. Commercial IR film was cut to fit the film holder, and signals were collected and then developed, all in total darkness. Exposure time in some cases was measured in hours. A densitometer was used to give a graph of intensity. Whilst tricky and tedious, the bonus was spectra with a high resolution of ~ 0.1 nm (1 Å). As seen in Fig. 1, it showed clear differences in signals between the electron and optical excitation. Many features would have been totally missed in modern systems that use bandwidths often as wide as 2 to 3 nm in order to boost both the intensity of very weak signals and rapid data collection. Higher resolution is invariably valuable, and a very recent luminescence example with nanoparticles of CaF2:Er indicated that high resolution (< 0.1 nm) revealed three sites instead of one [12].

Fig. 1
figure 1

Near-infrared spectra of a ruby crystal at 80 K recorded on infrared film with a bandwidth of ~ 0.1 nm (1 Å). There are major differences between electron and optical excitation. By contrast, modern equipment often employs bandwidth resolution over a wide spectral region (> 2 nm) (unpublished, Townsend 1961)

Full size image

One consequence of building and calibrating spectrometers was an awareness that there were potentially false signals from second- or third-order diffraction that require blocking filters. Intensity data also need careful calibration of both the wavelength and the grating sensitivity and detector response with wavelength. Unfortunately, these features are equipment sensitive and need cautious calibration (even for nominally the same models of the spectrometer and detector type as replacements may differ significantly). Further, simple software to transform data from wavelength (λ) to energy plots has often ignored intensity conversion from I(λ)dλ to I(E)dE, and hence, the energy plots are distorted and misinterpreted in terms of the number and shape of emission bands [13]. A common indicator of this transformation error is often a false “resolution” of several, progressively weaker, longer wavelength bands.

In addition to luminescence, a key feature of all defect studies of the 1950s was a strong focus on optical absorption that differentiated between various defect sites (extremely valuable as many sites have no associated luminescence). The next experimental advance was to use electron paramagnetic resonance analyses (EPR) as a probe of the extent of the site interactions with neighbouring ions. Less common was the further analysis using electron nuclear double resonance (ENDOR) to look at a larger scale environment. These methods were considered to be crucial, and hence, a majority of articles, texts, and books of that time include them. Indeed, many Watanabe articles rely not only on luminescence, but also EPR and optical absorption. By contrast, the majority of current articles associated with phosphor luminescence and TL rarely mention these more detailed and highly sensitive techniques. In part, this is because their sensitivity is extremely low compared with photon detectors. Nevertheless, they offer specific models as to local site environments. It is of course more difficult to derive absorption data with powder samples, as transmission may involve too much scatter, and reflectivity may include luminescence. Data for broad absorption bands may be more reliable than for line spectra from say rare earth dopants. This modern lack of absorption data is a major weakness if one wishes to discuss site models. There is therefore a real need to resolve these experimental difficulties.

An alternative view is that with an aim of a pragmatic improvement, and commercialisation of luminescence materials, then models of the sites are of minor consequence. Indeed, improvements can be tangible but a partial understanding of the relevant lattice sites may offer less significant benefits than empirical explorations of dopants and treatments (e.g. as in many TLD materials).

2 Changing Views on the Scale of Defect Interactions

Modelling and identification of lattice imperfections in insulators have advanced, but many of the established improvements in understanding are still being ignored [14]. The truly major interest in luminescence and optical absorption studies came from around 1936 when it was realised that a missing halide site in an alkali halide was charge compensated by trapping an electron. This “electron in a box” was a showcase success for quantum mechanics, as for all the alkali halides, the energy of the absorption band (F band) was closely predicted. The theory at that time could not cope with longer range excursions of the wavefunctions beyond the core site. Whilst later computation did advance, for experimentalists, there was a fixation that defect sites were tightly limited in range as “point defects,” and this has hampered not just early modelling of the sites, but it still exists in many examples of current literature, including luminescence and thermoluminescence. Equally, optical absorption, EPR, and ENDOR are now rarely attempted. However, they involve adjacent ions, so the “point defect” for 3 shells of neighbours implies ~ 27 interacting lattice sites. For many crystal structures, or natural minerals, the extra information is often too extensive for interpretation. The Watanabe papers with minerals sometimes emphasise this and he was clearly a leader in this respect.

Far less familiar is the luminescence analysis of exciton transitions at low temperatures which unequivocally were analysed not just in terms of electron-hole coupling at neighbouring sites, but distinguished separately data from every one of some 50 + neighbouring shells [9, 15]. These superb experimental data were recorded at 1.7 K and are of course thermally blurred and unresolvable by room temperature. Nevertheless, the interactions still exist, and this totally destroys the limited concept of very short-range interactions of isolated point defects. Unfortunately, we ignore this reality and tend to focus on the dominant short-range models of sites. Many phosphor and TL models assume that light production occurs when an electron is released from a trapping site, travels via the conduction band to a recombination site, and produces luminescence. Certainly, such events might occur, but the simplistic model misses the majority of events which involve structurally closely linked trap and recombination sites. This is despite the model that the historic TL material LiF TLD 100 is assumed to have an electron trap of a ring of 3 Mg2+ ions with 3 Li+ vacancies closely linked to Ti and O impurity ions that are involved in the emission site. Close coupling of the trap and emission sites guarantees very high TL efficiency, which would be less likely if it needed long-range transport.

A completely different demonstration of the presence and criticality of long-range interactions comes from crystal growth. For bismuth germanate (Bi4Ge3O12), there is a very complex phase diagram of some seven alternatives, so the growth of a single crystal is catastrophic as, during cooling from the melt, it produces a variety of phases, and the resulting sample has a dendritic appearance with scattering and poor optical quality. Skilled growers add just 2 ppm of a selected dopant that stabilises just one single phase. This produces excellent extremely large crystals. Vast quantities of them are then used in particle detectors for nuclear research. If a mere 2 ppm can stabilise the entire structure, then, if it were uniformly distributed, one dopant ion has influenced some 500,000 host ions. In some materials, current evidence suggests the RE ion dopants often enter the lattice as pairs or triplets, so the dopant site may be controlling a million host sites (i.e. a “box” volume of edge 100 ions). In fact, the interaction range must be greater as this assumes perfectly uniform rare earth distribution. Long-range interactions are clearly a major reality and totally opposite from the historic views of a very limited range of influence.

In many materials, associated trap/recombination pairs have been discussed. There is particularly clear evidence when using rare earth sites with identifiable spectra of the recombination site. Double doping can involve rare earth ions in roles of both trap and recombination sites [16,17,18]. Depending on the proximity of the two components, luminescence (and TL) can link them by excitation to higher states below the conduction band. This implies energy level-specific tunnelling between the ions. Hence, the TL peak temperatures will be energy level sensitive. Figure 2 offers an example of this via very low-temperature TL of Dy-doped zircon where the first three glow peaks appear at different temperatures depending on the energy levels of the rare earth (RE) recombination sites to which charge is transferred [19]. Long-range transport via the conduction band would be insensitive to the levels of the emission site. At higher temperatures, a Pr dopant in zircon illustrates more similarity in the peak temperature, but still with variations in the overall shapes as a function of the Pr transitions. Such wavelength sensitivity has been recorded in many examples of RE ions in other host materials, such as CaSO4 or CaF2.

Fig. 2
figure 2

Transition-dependent TL of Dy and Pr in zircon after X-ray irradiation [19]

Full size image

A historic feature was that size discussions were invariably in terms of ionic radii [e.g. 20]. This is a perfectly valid view, but for discussing defect sites, it is far better to consider the volumes of the various ions and additionally recognise that there may be alternative co-ordination numbers (CN) defining how many neighbours are associated [17, 18]. In the 1990s, the use of photon imaging tubes greatly enhanced TL spectroscopy and allowed low heating rates and high spectral resolution as exemplified by low-temperature TL of RE-doped LaF3 [21]. It was apparent that the TL peak temperature was linked to the dopant ion size, and also changed if there were different mixtures of dopants (i.e. such as Pr and Ho) at different concentrations (Fig. 3). Reassessing the data in terms of volume with a model of a triplet of CN6 RE dopants, occupying two La sites, offers a very simple linear pattern [14]. The triplet grouping is not the only option as one also sees an extension of the pattern with small ions that cluster in CN8 structures. Whilst the details are irrelevant here, the example emphasises how equipment improvements have opened opportunities that were previously unthinkable.

Fig. 3
figure 3

Peak temperatures at low temperature of RE-doped LaF3. The black data are for CN 6 size ions and the red for CN8. Two values are shown for Sm dopants at 0.1 and 1.0%. Values for undoped material differed between suppliers. A peak value for nominally pure LaF3 is indicated as La at ~ 128 K; however, somewhat purer samples had lower values down to ~ 122 K. The dashed trend line supports a model of three linked dopants, and perhaps surprisingly, it continues through the CN6 and the CN8 data to values seen in pure LaF3. The data in black (x) require a La displaced volume of less than a pair of La sites, whereas all the red data exceed this volume. The volume of a pair of La ions falls in the gap between the black and red data points (~ 12.5 nm 10−3)

Full size image

3 Interstitials

Returning to the very early alkali halide defect models, the optical absorption data offer a different optical absorption band attributed to halogen vacancies (F centre) and an interstitial with three halogens centred on a pair of halogen sites. The mechanism for this was proposed by both Pooley [22] and Hersh [23] which involved a replacement collision sequence along a <110> line of halogen ions. By cleaving a crystal in a UHV vacuum system and exposing the surface to exciton wavelength light, or an electron beam, not only were F and H centres formed but there was directional emission of halogens along both <110> and <121> directions which confirms the essence of the replacement collision formation model and the relatively long-range separation of a vacancy and an interstitial [24]. Unfortunately, later generations invariably state that when a vacancy is formed, then the displaced ion is the one that is the interstitial. This is incorrect, as for the alkali halides, the displaced ion is immediately next to the vacancy, and it initiates a replacement collision sequence along the halogen row and the H centre forms when the collision energy is dissipated (at an unspecified distance). In say KCl, both the immediate neighbour and the H centre are Cl. But in a KCl/KBr mixture, there is no reason why the displaced ion is the same as the extra ion in the H site. With optical absorption, there will be two types of F centre and two variants of the H centre, but one does not know if they directly correlate. Far worse is that the wording used in the modern defect literature for other materials invariably implies the displaced ion is precisely the one at the distant interstitial site, and this is definitely incorrect and misleading in the modelling.

Whilst this is not a new concept for movement enabled via a smaller size molecular state, it is clearly applicable to many materials. For example, in sapphire (Al2O3), creating a space between two oxygen ions as they form into a molecular state, or creating a chain of events along an oxygen direction, offers a low-energy route to separation of vacancies and interstitials. Alternatively, a transient small oxygen molecular state offers a gap through which other ions may pass. Lithium niobate has nominally the same structure as sapphire and whilst, in the excited molecular state, one can use an electric field to move metal ions (e.g. Li) into the structural vacancy site along the Z axis. This inverts the polarity of the crystal and can be done in a periodic pattern as needed for phase matching in second harmonic generation. Indeed, this concept was demonstrated successfully at a temperature well below the Curie temperature whilst electron irradiating a crystal that had an applied periodic electric field [25]. Such transient molecular states generate pathways that may well be operative in many crystal structures, but have rarely been considered.

4 Crystals, Minerals, and Powder

Natural minerals are inherently impure with a classic example of diamond that under natural formation processes is always contaminated with a very wide selection of ions other than carbon. Equally, modern phosphors excel because they contain a carefully selected range of dopants. Once we recognise the very long range of defect interactions, it is evident the attempts to enhance performance will always be predominantly an empirical process, with improvements by intelligent guesses at the probable types of defects that exist. Adding long-range effects to the modelling is difficult and generally overlooked. A further major assumption is that fundamental data acquired by studies of idealised single crystals (e.g. including optical absorption, EPR, ENDOR, etc.) is equally relevant to powder material. This is almost certainly misleading as the act of powder formation introduces dislocations, lattice distortions, very large reactive surface areas, and the opportunity for ingress of atmospheric dopants (e.g. water vapour etc.). Articles exist which indicate that the surface of powder grains is chemically different from the interior and dopants, such as rare earth ions used in phosphors, which can be located preferentially at surface and dislocation sites [26]. None of this is unexpected, but it adds more uncertainty in understanding the role of key sites within a lattice.

The use of high-resolution imaging technologies can identify dopants (such as RE ions) and indicate how they differ depending on whether one has a single crystal, a crushed powder, or different powder shapes. These opportunities have only been available for a few decades [26], and their importance has not fully been included in modelling of much of the luminescence literature. For example, with a high spectral resolution of Nd-doped CaF2, not only does the TL alter on moving from a single crystal to a crushed version of the sample [27], but later experiments revealed that in both cases, there are some 20 Nd line features and everyone differs from each other during TL. Simplistic models are clearly missing a wealth of information. This is not a unique example. However, it emphasises the vast diversity and complexity of the sites of imperfections.

5 Summary

The current group of papers recognises the very significant contributions of Professor Watanabe in his work related to defects involved in applications of luminescence and their analysis techniques. This particular contribution has attempted to highlight how the available experimental technologies and interpretive models have also dramatically altered during this extended period of time. In some aspects, it is clear we need to reintroduce some of his experimental procedures, such as optical absorption, EPR, and ENDOR. Also, we now benefit from more sensitive photon detectors with high-resolution spectroscopy and a wide range of imaging techniques. The aim of this overview is to emphasise both changes in techniques and concepts of defect structures and, in particular, highlight the very strong evidence for extremely long-range interactions that control the sites and their luminescence properties.