Metal-loaded organic scintillators for neutrino physics

The solvent molecules are the basis of each organic LS. Most solvent candidates are benzene derivatives, in particular alkyl benzenes like toluene, xylene or cumene, with an absorption band around 260 nm. Some basic properties of commonly used molecules are listed in table 1. The solvent of choice for high LY scintillators in past neutrino experiments was often pseudocumene (PC, 1,2,4-trimethylbenzene). However, over time safety considerations became more and more important and the focus shifted to high flash point solvents as linear alkyl benzene (LAB), phenyl xylylethane (PXE) or di-isopropyl naphthalene (DIN).

Table 1.  Density, flash point and the wavelengths of the optical absorption/emission peaks (dissolved in cyclohexane) for several solvent candidates are shown.

Molecule	Chemical formula	Density (kg l−1)	Flash point	abs. max.	em. max.
PC	C9H12	0.88	48 °C	267 nm	290 nm
toluene	C7H8	0.87	4 °C	262 nm	290 nm
Anisole	C7H8O	0.99	43 °C	271 nm	293 nm
LAB	—	0.87	$\sim 140\,^\circ {\rm{C}}$	260 nm	284 nm
DIN	C16H20	0.96	$\gt 140\,^\circ {\rm{C}}$	279 nm	338 nm
o-PXE	C16H18	0.99	167 °C	269 nm	290 nm
n-dodecane	C12H26	0.75	71 °C	—	—
Mineral oil	—	0.82–0.88	$\gt 130\,^\circ {\rm{C}}$	—	—

Material compatibility arguments sometimes exclude the use of pure PC or other chemically aggressive solvents. For example, when acrylic material is used as a detector vessel, the aromatic solvent is sometimes diluted in mineral oil (e.g. medicinal white oil) or n-alkanes, which can also help to raise the flash point of the scintillator cocktail. There can be the difficulty of electrostatic charging when the solvent is passing through pipes and filters, in particular when these materials are not conductive. Therefore, in some cases, to improve fire safety, an anti-static agent might be added to the solvent as e.g. Stadis-425 in the LS of the NOvA experiment [8].

A solvent candidate becoming very popular in the last decade is LAB, mainly due to its superior safety features, good material compatibility, high transparency and low cost. It was first proposed in the context of the SNO+ experiment [9]. On the other hand LAB based scintillators have reduced PSD capabilities and as a mixture of many isomers it does not have a well defined hydrogen or carbon fraction.

For some applications it is sufficient to have just one WLS (fluor) in the scintillator mixture. This dopand is typically added at concentrations of few g l⁻¹. An additional secondary WLS can be advantageous to improve the attenuation length of the scintillator. In particular if the detector volume is very large or the absorption band of the solvent molecule has a long tail towards longer wavelengths, an additional WLS is added at concentrations in the order of 20 mg l⁻¹ or less.

Table 2 shows the wavelengths of maximal absorption and emission for primary and secondary fluors. To minimize losses and ensure efficient energy transfer between the molecules, a large overlap between the emission spectrum of the donor molecule and the absorption band of the acceptor is mandatory. In general the absorption and emission spectra shift to longer wavelengths as more aromatic rings are added to the molecule.

A commonly used fluor combination is 2,5-Diphenyloxazole (PPO) and 4-bis-(2-Methyl-styryl)benzene (bis-MSB), although a better overlap of spectra could be reached using p-terphenyl (p-TP) with bis-MSB. However, p-TP has limited solubility in many solvents. Other alternatives for the primary fluor with very similar optical properties as PPO are 2-(4-biphenyl)-5-phenyl-1,3,4-oxadiazole (PBD) and 2-(4-biphenyl)-5-(4-tert-butyl-phenyl)-1,3,4-oxadiazole (butyl-PBD). The fluor 2-(4-biphenyl)-5-phenyloxazole (BPO) is known to provide high LYs, but on the other hand it is less favored due to its limited transparency in the region of scintillator emission and its toxicity.

As secondary WLS 1,4-bis(5-phenyloxazol-2-yl)-benzene (POPOP) can also be used, which has a slightly longer decay time than bis-MSB. In several liquid argon scintillation detectors tetraphenyl-butadiene (TPB) coated films are preferred for efficient light conversion from the argon emission (128 nm) to the blue-visible region (400–500 nm), where photomultiplier tubes (PMTs) are most sensitive. An exceptional large Stokes shift, the difference between positions of the absorption and emission peak, is observed in 1-phenyl-3-mesityl-2-pyrazoline (PMP). The PMP Stokes shift is about twice as large than in other fluor candidates.

A common feature of all WLS is the high fluorescence QY, the probability to reemit a photon after absorption. When diluted in an inert medium as cyclohexane the QY of basically all molecules of table 2 is above 80% [10, 11]. However, there can be strong concentration and solvent effects changing the QY numbers significantly. Moreover, for many fluors there are contradictory results in the literature values of this quantity [12].

Volatile components dissolved in the liquid as oxygen or radioactive noble gases are removed by gas, typically nitrogen, stripping. The method is based on differences in the equilibrium composition between liquid and vapor. Approximations on the purification factor can be made by use of Henry's Law. The technique is very effective in particular by applying counter current flow of the liquid and the gas. The efficiency can be further increased by using columns with structured packing [20] or higher temperature, implying reduced solubility of gas contaminants. To avoid impurities such as radon or krypton being added by the gas purging, the purification and selection of clean nitrogen is essential [21].

Distillation is an effective process to improve the scintillator transparency and to reduce radioactive impurities in the solvent liquid such as daughters of the radon decay chain [22]. All impurities which are less or more volatile than the solvent can be separated. Since the more volatile components are effectively removed by gas stripping, the focus is usually more on the less volatile ones. When distillation is applied to the full scintillator mixture one has to be careful not to separate the solvent from the fluor [20], since both of them have very different boiling points. Scintillator solvents are often extracted from oil products and separated from other oil components by distillation at the supply company. Mono-molecular solvents such as PC can be made intrinsically purer as mixtures, since the distillation parameter ranges for temperature and pressure can be set narrower.

In particular, multi-staged vacuum distillation [23] is applied as a purification step for LAB and PC. The vacuum distillation column should be composed of many theoretical plates that can effectively separate the organic solvents with different boiling points, e.g. at $\sim 10\,^\circ {\rm{C}}$ intervals. Typically, distillation is set to finish when more than 90% of the solvent is collected. The impurities could be for example partially oxidized organic molecules having lower volatility compared to the pure solvent. Distillation is one of the best methods for improving the optical transmission of a LS solvent [23]. Optical purity is essential for high light collection and to achieve good energy resolution in a neutrino detector.

In the water extraction method impurities in the organic LS are transferred into an immiscible aqueous phase based on their relative solubilities. An advantage of water extraction over distillation is the ability to directly process the full scintillator mixture allowing for on-line repurification of the detector liquid. This method is highly effective at removing polar or charged substances. It significantly lowers the amount of radioactive impurities such as ²³⁸U, ²³²Th or ⁴⁰K. Most of the radioimpurities typically enter the scintillator mixture via the primary fluor. In particular the potassium (K) content in a primary fluor as PPO can be in the ppm level, far above the usual specifications of a low background detector. Water extraction in concentrated PPO solutions was found to be a promising way of removing K [24]. On the other hand, water extraction is less suited reducing optical impurities, which are mainly of organic type and would require phase separation from the scintillator solvent.

Traces of chemical impurities can be removed from the scintillator by sending the liquid through a solid column of an adsorber material as aluminum oxide, silica gel or an other molecular sieve. A prerequisite for a good adsorber material is a large surface-to-volume ratio. Other important properties are adsorber parameters such as pore size and surface conditions. Many adsorber materials are hygroscopic and should be dried at elevated temperature before use to remove water molecules blocking adsorption sites. The ratio of impurities on the adsorber surface and in the liquid is determined by a specific equilibrium constant. The equilibrium constant depends on the free enthalpy of the system and the temperature. The free enthalpy is constant for a given solvent, impurity and adsorber type.

At the beginning of the purification process the impurity concentration forms a gradient along the column from low purity at the entrance of the column and high purity at the end. At some point the impurity is travelling through the column and reaches the exit, the so-called breakthrough. In a dynamic situation of a constant flow, the number of segments in a column depend on the time constant to reach the equilibrium situation and the velocity of the liquid through the column.

A weak acidic alumina column was found to significantly improve the attenuation length of scintillator solvents as e.g. o-PXE [25]. The purification normally has no significant effect on the scintillation yield. Since the chemical composition of most of the different impurities that are removed by the column are not well defined, several layers of activated column packings differing in pH can be used [26]. In this multi-stage column approach acids, bases as well as neutralized particles can be effectively removed.

The emission spectrum of a LS is determined by the fluors, since the de-excitation of the solvent molecules is mainly non-radiative. To measure the characteristic emission spectrum of a fluor, the molecule is diluted in an optically and chemically inert solvent as cyclohexane. However, in the scintillator solvent the spectrum can be shifted by up to 10 nm compared to isolated molecules. The interaction with neighboring solvent molecules might affect the levels of the electronic states especially for polar substances. In addition there can be concentration effects, in particular for the primary fluor which is typically added at few g l⁻¹, which might cause a distortion of the spectrum.

As the light propagates through the liquid in a large scale detector, it can be absorbed and re-emitted by the fluor. Therefore the emission spectrum changes and shifts to longer wavelengths, since the photons are only re-emitted at equal or lower energies.

To simulate the detector response it is important to know the primary scintillation spectrum after non-radiative energy transfer including all solvent and concentration effects. To avoid shifts of the spectrum due to intrinsic absorption and reemssion processes, the primary light emission should be detected at the same side of the cell where excitation takes place ('front face geometry'). Then an optical model should be applied to take care of the light propagation in the detector from the point of production to the photosensor. The model should include the most relevant parameters as absorption properties of the components, re-emission spectra and probabilities, QY of the fluors and scattering cross sections.

The intensity I of a one-dimensional monochromatic light beam propagating through a LS decreases exponentially with the attenuation length Λ. If the attenuation length is measured using a standard UV photospectrometer, light can be lost either by absorption or by scattering. The absorbance A after a distance x is defined as $A(x)={\mathrm{log}}_{10}(I(0)/I(x))$. The total absorbance can be calculated by summing up the absorbance contributions of the different LS components. Accordingly, the overall attenuation length ${{\rm{\Lambda }}}_{{LS}}$ is obtained from the single contributions ${{\rm{\Lambda }}}_{i}$ via

$$\begin{eqnarray}&&\displaystyle \frac{1}{{{\rm{\Lambda }}}_{{LS}}}=\displaystyle \frac{1}{{{\rm{\Lambda }}}_{1}}+\displaystyle \frac{1}{{{\rm{\Lambda }}}_{2}}+\cdots .\end{eqnarray} \tag{ 6 }$$

Here the different terms can represent different processes as absorption or scattering. The formula can also be used in case the ${{\rm{\Lambda }}}_{i}$ are determined separately for the different components in the mixture from individual measurements.

The linearity between absorbance and concentration is known as the Beer–Lambert Law. The validity of this Law is in principle constrained to low concentrations. However, if we look at wavelength regions in the tail of the spectra the absorbance is mainly on impurities of the sample, which are low in concentration. Therefore, equation (6) can be used in the region of LS emission which is mostly in the high wavelength tail of the spectra.

In the wavelength region of UV light below 280 nm the solvent molecules dominate the absorption in a LS as illustrated in figure 2. At higher wavelengths, in the region of solvent emission around 300 nm, the main absorption is on the primary fluor, typically up to about 350 nm. The secondary fluor, if present, dominates absorption in the region of the scintillation light seen by the photomultipliers, which is normally above 350 nm. The molar extinction coefficient of secondary fluors like bis-MSB are dropping steeply above 400 nm and the absorbance starts to be negligible above 430 nm [26]. In this optical region the impurity levels in the chemicals and scattering processes are most relevant for the attenuation length.

Zoom In Zoom Out Reset image size

Highest attenuation lengths for solvents are typically found in pure n-alkanes as n-dodecane. As delivered from companies it can be already more than 10 m at wavelengths around 350 nm [26]. In typical cases the attenuation length is monotonically increasing from 350–580 nm. Attenuation lengths of more than 10 m can also be found in LAB [27]. For some LAB samples however, there can be wiggles in the absorption spectrum in the region between 350–400 nm possibly caused by contaminations of bi-phenyls. In such a case a secondary fluor should be used. Pseudocumene has a slightly lower attenuation length above 400 nm than LAB or n-alkanes. For PXE or DIN the attenuation length is typically even lower, in a range of 1–5 m around 430 nm. However, the lower values of PC and PXE are partly due to a reduced scattering length compared to LAB or oil [28]. With alumina oxide column purification the attenuation could be improved to about 10 m for PXE [25] and to about 4 m for DIN [29].

For high purity PPO the attenuation length in a transparent solvent at concentrations of few g l⁻¹ is still well above 10 m [26]. In general, for solvents as well as fluors, there are significant batch to batch variations even for material from the same supplier and identical quality specifications. Therefore, careful material selection, testing, liquid handling as well as close contact to the companies are mandatory to reach the highest optical purity. As an example the attenuation length of purified PXE with and without fluors is shown in figure 3.

Zoom In Zoom Out Reset image size

The nominal attenuation length of LS (after purification) can be longer than 10 m. The typical 10 cm pathlength UV spectrometry, which is adequate for long-term optical stability monitoring at 5% variation, has limited precision for absolute measurements of the attenuation length at very high transmission. To further reduce the systematic errors, long-arm attenuation systems are essential.

In a specific setup [27], a 2 m long stainless steel tube (inner) coated with Teflon is vertically aligned and contains several optical components in addition to a PMT. Such a system is optimized to measure long attenuation length LS. A cylindrical quartz tube (2.5 cm diameter) is connected to the steel tube where the liquid is contained for pathlength control. The vertical orientation allows the selection of various pathlengths in the liquid from a few centimeters to 2 m, in order to measure the variation of transmitted light as a function of propagation distance in the liquid. Pulsed LED beams, at selected wavelengths from 350 to 500 nm, are transmitted by an optical fiber to a beam-splitting cube that produces two beams with 50/50 intensity ratio. The different wavelengths give the capability to investigate the Rayleigh scattering, which has a sensitive dependence on wavelength. The split beams are collected by fiber collimators and transmitted again by 2 m and 40 m optical fibers, respectively, to give a time delay ∼200 ns between the two pulses. One beam measures the absorption of the scintillator, while the other measures the variability of the pulsed LED output. Both light beams are collected by the same PMT and the signals are analyzed. Such a system gives a sensitivity of absorption length measurement at $\gt 30\,{\rm{m}}$ [27].

The transparency of LSs, whether they are metal loaded or metal free, is very sensitive to the liquid handling and storage conditions. The attenuation length of a pure PC solvent can reduce by a factor of more than three within few weeks when stored at room temperature in an open steel vessel [20]. Exposure of scintillator components to air can lead to oxidized products absorbing light in the region of scintillator emission. Materials such as stainless steel can act as a catalyst for the reactions and contact to the organic liquid might reduce the optical stability. To minimize such yellowing an antioxidant as Vitamin E or BHT might be added to the solution [8]. It was found that Vitamin E is also fluorescent and can contribute to the scintillation [30].

The attenuation length in a one-dimensional beam is limited by scattering. Whereas absorbed light with no re-emission is lost in a large scale LS detector, scattered light just changes propagation direction. Therefore, scattered or re-emitted light could still be detected at the photosensors of a three-dimensional detector.

In a pure scintillator mixture Rayleigh scattering is dominating over Raman or Mie scattering. Without absorption, considering Rayleigh scattering alone the attenuation length at 430 nm for standard scintillator solvents is in the range 20–50 m [28]. This means for attenuation lengths of more than 20 m, as measured in a UV/Vis spectrometer, more light is scattered than absorbed without re-emission. In water based LS scattering can be significantly higher compared to pure organic liquids due to the organic/water interfaces. There is a strong wavelength dependence for Rayleigh scattering. The scattering length rises as ${\lambda }^{4}$. Although light scattering has a small impact on the energy resolution of a detector it has the negative aspect of deteriorating the capabilities of time or vertex reconstruction.

In a two component scintillator system, for low fluor concentrations, the LY strongly increases with increasing fluor loadings. Energy transfer from the solvent to the fluor competes with internal losses of the solvents. About half of the maximal LY is reached at the critical concentration [31]. This critical concentration depends on the solvent/fluor combinations and is lower if the non-radiative transfer is effective. The critical concentrations for standard scintillators is typically less than 1 g l⁻¹.

Above the critical concentrations the LY keeps increasing with increasing fluor concentrations, but the curve flattens until it reaches saturation around 5–10 g l⁻¹. At more than 10 g l⁻¹ the LY starts to lower again due to self-quenching. Self-quenching is an effect related to interactions between unexcited and excited fluor molecules in which the excitation energy can be lost by collision. The LY dependence on the fluor concentration is illustrated in figure 4 for the case of BPO in methoxybenzene (anisole) [31].

Zoom In Zoom Out Reset image size

Absolute measurements of the scintillation photons emitted for a given energy deposition in a scintillator liquid are difficult. Therefore most measurements are done relative to a reference scintillator with a known LY. To minimize the systematic uncertainty samples and reference should be measured in the same setup. Typical reference samples of use are anthracene or the commercial scintillator BC-505 from Bicron company (St. Gobain Crystals). In the data sheet provided by Bicron, the specified LY of the BC-505 scintillator is 80% anthracene. However, the literature values for the absolute LY value for the anthracene itself (in photons per MeV) vary up to 25%.

Highest LYs of about 13000 photons/MeV are obtained with PC based scintillators. For Borexino, which is using pure PC as solvent, the yield is a little bit lower, 11500 photons/MeV [32], since the PPO concentration is rather low (1.5 g l⁻¹). In KamLAND the PC is diluted in n-dodecane. The aromatic fraction is at 20%, nevertheless the LY is still more than 80% of the Borexino value at the same PPO concentration [33]. Even for pure mineral oil with a fluor there is some scintillation light. The amount depends on the exact chemical composition and in particular the aromatic fraction of the oil. The LY with pure oil as solvent is usually less than 30% of pure PC [34]. But even for a pure n-alkane significant amount of scintillation light can be found [31]. In this case without aromatic ring structures and no π electrons, the excited state is populated by processes like ionization or free radical formation with subsequent recombination.

In the Borexino CTF [24], a PXE based scintillator was tested with p-TP as primary and bis-MSB as secondary fluor. The LY was measured to be 93% of the Borexino mixture [25]. Typical values for LAB based scintillators are slightly below the LY found in the KamLAND mixture. The Daya Bay experiment found similar values for LY and attenuation length in their Gd-loaded and unloaded LAB based LS [35]. In table 3 those numbers along with some other properties of several LS are listed.

In case of metal loadings an additional energy transfer path opens from the solvent molecules to the metal compound which competes with the transfer to the fluor. In the scintillator design this transfer to the metal material has to be minimized compared to the transfer to the fluor, since the metal compound in general is non-fluorescent. Therefore, light transferred to the metal compound is lost for the scintillation process and reduces the LY (quenching). The LY in metal loaded LS can be improved by an increase of the primary fluor concentration [31], which strengthens the energy transfer path from the solvent to the fluor, or by adding an additional WLS. However, at too high fluor concentrations the light loss due to the self-quenching effect will exceed the light gain due to more effective energy transfer from the solvent to the fluor.

The decay time in a LS can have several components. The lifetimes of the different molecular excited states in the scintillator medium determine the pulse shape of the events. Therefore, experiments using pulse shape analysis to discriminate their background strongly depend on the values of their fluorescence decay-time constants. The time behavior of the scintillator emission is often described by a multiexponential function with three or more decay constants.

The shortest decay constant (few ns) is attributed to the fluorescence decay time of the primary fluor, but it also includes the time for energy transfer from the solvent to the fluor, which might occur in several steps. Therefore, with increasing fluor concentration the decay time gets faster (more direct fluor excitation) until the decay constant approaches the intrinsic value of the isolated fluor molecule. This shortest decay constant is the dominant contribution to the decay profile. The longer components could arise from de-excitation of triplet states or other rare processes. Scintillator mixtures based on oil admixtures or LAB have longer decay constants than solvents as PC or PXE [36]. The effect of adding a secondary fluor as bisMSB to a mixture containing PPO has only a small effect on the decay times.

At low energies up to about 200 keV, the background in an organic LS detector is limited by the β-decay of intrinsic ¹⁴C (${E}_{\max }=156$ keV), a component which can not be chemically separated from the liquid. In the atmosphere the ratio of ¹⁴C/¹²C is about 10⁻¹² g g⁻¹. The production mechanism of ¹⁴C is cosmic ray interactions on nitrogen. In petrochemical organics, which are stored underground for millions of years, most of the original ¹⁴C (half-live of 5730 years) has decayed. In this way, the isotopic fraction can be reduced up to 6 orders of magnitude [37].

To reach and maintain high radiopurity in a LS, contact of the liquid with dust (U, Th and K contaminations) and air (Kr and Rn) have to be avoided. In Borexino [18] and KamLAND [19] radiopurity levels of 10⁻¹⁷ g g⁻¹ and less for ²³⁸U and ²³²Th were reported. It was found to be hard to remove the alpha emitter ²¹⁰Po, a daughter isotope of the ²³⁸U chain. It is generated via β-decay of ²¹⁰Bi, which itself is created by the long-lived ²¹⁰Pb (half live of 22.3 years). Alpha particles can be discriminated in LS with good particle identification via pulse shape analysis or they are even below energy threshold because of ionization quenching. Therefore, the contribution of ²¹⁰Bi might be more critical as compared to ²¹⁰Po in terms of background [18, 38]. Those isotopes at the end of the U decay series are often not in secular equilibrium with the rest of the chain. They can be supplied from the surface of the detector vessel and might not be intrinsic impurities of the LS itself [38]. The equilibrium can also be broken at the level of radium or the mobile and omnipresent noble gas radon. For ⁴⁰K levels of less than 10⁻¹⁶ g g⁻¹ could be reached. To reach the purity requirements in the solutes, in particular for K, U and Th, is typically more challenging than for the solvents.

Special emphasis in scintillator purification is on radioactive isotopes of noble gases. Those are removed by gas stripping using ultrapure nitrogen [21]. In the nitrogen plant emanation of Rn isopotes from materials in contact with the gas have to be minimized. In some applications, e.g. in Borexino, purity levels of less than one ²²²Rn atom per m³ of LS are required and were achieved [37]. Reaching and maintaining such radiopurity levels is even more difficult when the LS is loaded with metals or other elements.

The loading of metal carboxylates into scintillator can be achieved either by liquid–liquid extraction in a two phase system of an organic LS and an aqueous phase or alternatively by solid dissolution. In the first approach, a large quantity of metal carboxylates can be solvent-extracted into the LS directly in one step within a few hours. The solid-dissolution approach needs first preparing the metal-carboxylate solids, followed by dissolving it in the organic scintillator. Although the process of solidifying the compounds involves several steps that could enhance the probability of introducing impurities during the preparation, with careful control of production procedures, both methods give similar properties of metal-doped LS in terms of LY, optical transmission, and chemical stability. The synthesis of metal carboxylates, such as indium, ytterbium, and gadolinium, as neutrino- or neutron-capture targets in LS was developed through studies for solar neutrino experiments [54, 55]. Both ytterbium and indium can be doped into PC or LAB as high as 8–10 wt%. The implication of gadolinium carboxylates was further focusing on the reactor antineutrino oscillation experiments, such as Palo Verde, Daya Bay, and RENO, at 0.1%–0.2%.

The requirement of purity for the metal-doped scintillator depends on the physics goals of the experiments. Special cares are often given to the inherent radioactivity, radon prevention, optical cleanness and chemical quantitative analysis, particularly the metal/H and C/H ratios. Most of the purification steps have to be applied before and during the synthesis of the metal carboxylates. Chemical separation schemes that would be used after the metal-doped scintillator has been synthesized are unsuitable because they would likely remove some of the metallic ions along with the other inorganic impurities.

The developed purification techniques are divided into two categories: the removal of chemical and of radioactive impurities. The levels of non-radioactive chemical species that can adversely affect the optical property of the scintillator have to be strictly controlled. Indeed, the removal of chemical impurities not only increases the light transmission in the scintillator, but also enhances the long-term stability of the metal-doped LS. Some impurities could induce slow chemical reactions that would cause the transparency of scintillator to deteriorate gradually.

Note that the quality and impurities of the carboxylate scintillator can vary with different vendors even at the same quoted purity level, as previously mentioned. Thus the properties of scintillator, such as its metal-loading capability, photon production yield and optical property, might change from one vendor to another and even from one production batch to another. To avoid these problems, several chemical purification steps are developed for quality assurance during the synthesis phase:

• (1)  
  The purification of chemical ingredients in the aqueous phase, including ammonium hydroxide and ammonium carboxylate, can be done by solvent extraction with toluene mixed with 0.5% by weight of TBPO.
• (2)  
  The low-volatility organic solvent, LAB or PC, is purified by dry column absorption (e.g. 50-cm long glass column loaded with activated neutral Al₂O₃ of 150 mesh, ∼58 A particle size) or multi-staged vacuum distillation (see section 2.5.2).
• (3)  
  The high-volatility or viscous liquids, such as the carboxylic acids, can be purified by temperature-dependent, thin-film vacuum distillatory using a Teflon wiper for layer formation of inlet solvent within the distillation column and a mechanical rotary vacuum pump to keep the system pressure $\lt 0.01\,{\rm{bar}}$.

Different procedures have to be used for inorganic species such as Gd³⁺ and other lanthanides, which often contain naturally occurring radioactive impurities, such as Th⁴⁺, U⁴⁺ and U⁶⁺ (as ${\mathrm{UO}}_{2}^{2+}$). Purification steps will have to be developed to reduce these radioactive species to the concentration of requirement (i.e. 10⁻¹² g g⁻¹ in the 0.1% Gd-LS for Daya Bay). Conventional purification methods are cation exchange (i.e. Chelex-100) and solvent extraction such as TBP, TOPO, TBPO, and TTA used in chemical processing in the nuclear industry. However, both techniques require large chemicals or exchange resins that would require special chemical handling and are not cost-effective. A new self-scavenging method utilizing different metal-hydrolysis processes by pH-control, followed by micro-filtration (2 micron) is developed to purify gadolinium from its lanthanide and actinide impurities. In this study, thorium and its decay daughters are completely removed from Gd at $\mathrm{pH}\gt 6$ [56].

The carboxylic acids often exist in dimer forms and could be deteriorated by photo-induced oxidation in the presence of oxygen. A freshly purified batch of carboxylic acid should be prepared within 24 h before the metal-complexing synthesis. On the other hand, after interaction with metallic ions, such optical degradation is greatly reduced due to the formation of metal-oxygen bonds. In the reactions with indium or gadolinium, the metal carboxylates show no severe degradation effect as does pure carboxylic acid. Yet, for precaution, keeping them in the dark under a nitrogen environment is highly preferred, not only to extend the shelf-life of metal-carboxylate LS, but to avoid the oxygen quenching that could reduce the LY by 15%–20% for air-saturated scintillators.

The scintillator emission spectrum is adjusted using WLSs to the most sensitive region of the PMTs, typically to a wavelength range around 420 to 450 nm. The dissolution of purified metal carboxylate at few tenths of percent (i.e. Gd at 0.1 wt%) does not significantly affect the transmission of scintillation emission in this wavelength region.

The development of loading metal carboxylates in either singular or binary scintillator systems is mature. The metal carboxylates can be dissolved up to several weight percent in a variety of scintillators including PC, PXE, PCH, LAB or their mixtures with n-dodecane or mineral oil. A very consistent procedure, with $\gt 95$% Gd chemical extraction efficiency for preparing 0.1%–0.2% Gd-loaded PC or LAB has been established for reactor antineutrino experiments [35, 49]. Ideally the nature of scintillation photon yield is a function of weight fraction of fluors in the scintillator. A scintillation cocktail of linear alkyl benzene (e.g. Cepas) with the wavelength-shifting fluors of PPO (3 g l⁻¹) and bis-MSB (15 mg l⁻¹) gives 9200–9500 photons/MeV depending on the vendor and method of preparation. Carboxylate is a self-quencher that could affect the FRET between fluors in the LS and should be controlled at minimal level, just enough for complexing metallic ions in the LS. For instance, 0.3 wt% of carboxylic acid in 0.1 wt% Gd-doped LS gives about 90% LY relative to the metal-free LS version.

Based on LENS R&D activities on Yb, In and Gd-LS by an INFN(Italy)/INR(Russia) group [48, 52, 57], a Gd-LS was produced for supernovae neutrino detection in the LVD experiment at LNGS, Italy [58]. Two modules of the LVD detector, each of them containing 1.2 tons of LS inside a 1.5 m³ stainless steel tank, were loaded with about 1 g l⁻¹ Gd to increase the signal to noise ratio for ${\bar{\nu }}_{e}$ detection. The unloaded LS in the LVD experiment is a mixture of aliphatic and aromatic (8% or 16%) hydrocarbons known as White Spirit or Ragia Minerale. The light attenuation length of this liquid at 425 nm is well above 10 m. The Gd compound used was prepared as solid Gd-2-Methylvalerate salt following the procedure developed in [57]. A 15 kg batch could be produced in one month using standard equipment of a chemical laboratory. The salt was dissolved in a highly concentrated solution and subsequently diluted in the tank. The stability and performance of the 2.4 tons of Gd-LS were tested. Attenuation length after Gd-loading was still above 10 m in one of the tank. For the other tank a reduced transmittance with an attenuation length of about 6 m was observed directly after loading. From periodical measurements in an UV/Vis photospectrometer over three years stable behavior of the attenuation length was observed in both tanks. A degradation of the absorbance of more than 1%/year could be excluded from the measurements.

The remarkable stability of BDK metal complexes is mainly due to the strong chemical bond of the BDK anions to the centrally chelated metal ion. In this way, the metal gets trapped in a stable organic structure and creates a compound that can be dissolved in organic LS. The simplest form of a β-diketonate is acetylacetone (C₅H₈O₂, 2,4-pentanedione, commonly Hacac). The In form of this BDK version (In(acac)₃) was successfully tested in the LENS low background facility (LLBF) at the Gran Sasso underground laboratory in Italy [65]. The Hacac was mainly chosen as ligand to the In metal because of the low molecular weight. The lower the weight of the non-fluorescent metal compound, the higher the weight of the components participating in light production in the LS.

In many applications the metal to be loaded to the scintillator is a rare Earth as gadolinium (reactor neutrinos), neodynium ($\beta \beta$-decays) or ytterbium (solar neutrinos). Other than In(acac)₃, trivalent rare Earth acetylacetones tend to form hydrates (M(acac)₃ · nH₂O). The literature indicates that basically only two hydrates are formed: the monohydrate (n = 1) and the trihydrate (n = 3) [66]. The trihydrates is the most common form after synthesis. From that molecule the monohydrates can be produced by drying methods or by crystallization from a solution of the trihydrate in e.g. ethyl alcohol or acetone. The hydrates of the rare Earth elements with lower atomic number are more stable. The larger diameter of the latter probably enable them to accomodate more easily the additional water molecules.

The removal of the last water molecule in the monohydrates might be achieved under vacuum at elevated temperature of about 60 °C [67]. However, the material could decompose in this step and partially be converted to M(acac)₂OH which is reported to be stable for Gd [68]. To avoid such complications and to improve the solubility in organic solvents the Hacac ligand can be replaced by bigger BDK molecules as Hthd (C₁₁H₂₀O₂, 2,2,6,6-Tetramethylheptane-3,5-dione). This more stable form was chosen e.g. for the reactor neutrino experiments Double Chooz [51] and Nucifer [63].

The solubility of the BDK metal complexes in organic solvents might depend on the crystal structure of the powder material. In general, correlation between high vapor pressure and high solubility can be found [69]. For Gd(thd)₃ high solubility is found in ethanol and very high solubility in tetrahydrofurane (THF). The latter can be used to break up the crystal structure of the powder and its presence in the scintillator has a stabilizing effect on the liquid. One or more THF molecules can stick to the complex when the coordination number of the central metal atom is above six. In this way it protects the complex from attacks of other molecules as water, which could lead to decomposition. In most complexes volatility, thermal stability and solubility are increased when fluorine substituents are present in the ligand shell [69]. An example for such a BDK molecule is heptafluorodimethyloctanedione (C₁₀H₁₁F₇O₂, Hfod), which on the other hand is typically limited in terms of transparency.

In the synthesis of the metal BDK molecules the starting material is typically metal chloride, oxide or nitrate. Below the basic synthesis steps for two different metal BDK systems, In(acac)₃ and Gd(thd)₃ are described. For the production of In(acac)₃ as described in [50], InCl₃ is first dissolved in water at a pH around 2 and Hacac is added to the solution. Then the indium material is transferred into an organic phase by solvent extraction. For this an organic solvent such as 1,2-dichlorobenzene, which has high solubility for In(acac)₃, is added and the solution is neutralized using a base (NaOH solution). One of the main challenges in the synthesis is to avoid hydrolysis as much as possible and the removal of hydrolysis products. After approximately one day the organic phase containing the indium is separated from the aqueous layer. The solvent is removed under vacuum at elevated temperature resulting in the formation of white In(acac)₃ crystals. The crystals can be purified by washing them in 2-propanol in which the solubility of In(acac)₃ is negligible. The final material is obtained after filtering and drying and can be dissolved directly in the LS, typically up to indium concentrations of around 1 wt% and above.

A way to synthesize trivalent rare Earth complexes as Gd(thd)₃ with yields of more than 90% is described in [60]. The H(thd) is first dissolved in ethanol and connected to a vacuum system. To reduce impurities in the H(thd) it might be distilled before usage. Then a solution of NaOH dissolved in a 50% ethanol/water mixture is added while stirring. The metal nitrate, also dissolved in 50% ethanol/water, is added to the reaction flask. The solution in the flask is stirred under vacuum and its volume reduced by evaporation of the ethanol. The metal complex precipitates after the addition of pure water. The crystals get isolated by vacuum filtration and are dried after washing with water and/or a water/ethanol mixture. As a final step the material should be sublimed to reach highest purity of the crystals.

The stable structure of metal BDK complexes allows heating of the compound without decomposition and in combination with the typically high vapor pressures purification steps by sublimation can be applied. For the sublimation the sample material is placed into a glass tube. The dimensions of the tube depend on the sublimation rate to be achieved. A vacuum is applied and the metal BDK complex is heated, typically to around 200 °C, using a heating mantle. The vacuum is drawn at the exit of the tube while a slight flow of N₂ or argon carrier gas enters from the upstream side. The vapor condenses downstream in an unheated region and pure crystals start to grow [50]. With this technique purification rates up to 1 kg d⁻¹ could be achieved [26].

In the sublimation step the complex gets separated from impurities as partially hydrolyzed components or BDK complexes with other central metal ions due to different sublimation parameters. Furthermore, sublimation can improve significantly the radiopurity of the sample. In some samples the potassium contamination was found to be more than two orders of magnitudes lower for the sublimed compared to unsublimed material [26]. The absorbance of the powder in the wavelength region of interest can be reduced in the sublimation step by a factor 2 and more depending on the purity of the starting material. Molar extinction coefficients well below 0.01 l (mol cm)⁻¹ above 420 nm were reached [26, 50]. At such purity levels the transparency of the LS is typically dominated by the primary fluor and several meters of attenuation length can be achieved even with metal loadings of few per cent. The influence of the metal BDK on the transparency of the LS is usually weak and less critical than the impact on the LY caused by quenching effects.

The addition of a metal complex to the LS opens an additional energy transfer path from the solvent molecules to the typically non-fluorescent metal molecule. This transfer is in competition with the transfer to the primary fluor, which is relevant for scintillation light production. The energy transfer to the metal BDK molecule can be rather effective, since the absorption spectra are in similar wavelength regions than the one of the primary fluor, around 300 nm [70]. For the same molar concentration of primary fluor and metal complex the energy is sometimes even more likely transferred to the metal compound than to the fluor [31]. This means for reasonable LYs (>50% of the unloaded scintillator) the fluor concentrations should be about as high as the metal concentration. However, at high metal and fluor loadings self-quenching and self-absorption of the primary fluor become significant.

Organic LS dissolved at sufficient concentration in a water medium is essential to the success of water-based LS. Chemically organic solvents are immiscible in water mainly due to the differences in polarities. It will entail a surfactant that contains lypophilic and hydrophilic groups to emulsify the organic LS into the water solvent. Engineering of a complexing medium to stabilize the lypophilic and hydrophilic molecules in a water medium with appropriate optical transmission and long-term stability is essential. The solubility of a series of conventional organic LSs, such as PC, PXE, PCH, LAB, commercially capable in large quantity, have been systematically studied in water. Their loading capabilities, optical properties, purification methods, proton densities and commercial availabilities are well understood. A suitable ampliphilic surfactant or mixtures of two surfactants that can reduce the liquid–liquid interfacial tension between organic solvent and water, and thus blend the LS into the bulk water molecules, are necessary to stabilize the WbLS solution. The degree of tension reduction is a function of surfactant concentration, which could also affect the optical and stability properties of the medium.

It should be noted that loading of a WLS into water is different from the water-based LS. The former only improves the Cerenkov counting efficiency; but has no capability of capturing incident UV photons induced by ionization radiation. Linear alkyl benzene currently used in several neutrino experiments (Daya Bay, SNO+, Reno, JUNO) is the scintillator solvent of favor for WbLS production. LAB has a long alkyl chain, 10–13 carbon atoms, attached to the benzene ring (where the unsaturated π-electrons are essential for emission of ultra-violet light due to incident ionizing radiation). The worldwide industrial uses of LAB (millions of tons per year) are for lubricants and organic solvents in application of manufacturing detergents, in which a sulfonic-acid (HO–S–O–OH) (LAS) or amine derivative (PRS) of alkyl benzene is produced. Briefly, the chemical basis of linear alkylbenzene sulfonate derivatives is that the alkyl chain and benzene group act as a solvent to extract oils and grease (lypophilic), while the attached sulfonate or amine group is soluble in water (hydrophilic). The unique lypophilic and hydrophilic properties, and associated benzene π-electron ring, acting as surfactant and LS simultaneously enable either PRS or LAS to be a promising anionic surfactant for preparing a water-based LS (or organo-scintillator loaded water detector). Similar results can also be achieved by non-ionic surfactants that have a hydrophilic polyethylene oxide chain (on average of 9.5 ethylene oxide units) attached to an aromatic hydrocarbon lipophilic or hydrophobic group (NP-4,-40, TX-100, 114, 405, etc). The production quantity and cost of these surfactants are very attractive for large-scale neutrino detectors.

New water-based LS R&D based on these surfactants was initiated by the BNL neutrino and nuclear chemistry group in 2010. A series of WbLS samples loaded with different percentange of organic scintillator in water in use for different physics applications have been successfully synthesized and stable for years since preparation. A > kt scintillation water detector [77] capable of Cerenkov and scintillation detections needs long attenuation length at tens of meters in the PMT sensible region. In the example of lightly organic doped (1 wt%) water, its absorption length is superior to conventional pure LS detectors; however the main challenge of optical transmission is the scattering effect introduced from the inhomogeneous charge distribution between the water and organic interfaces. A mixture of non-ionic and ionic surfactants could improve the scattering significantly. The WbLS principal could also be applied to loading of metallic ions directly in aqueous solution to provide sensitivity for a wide range of particle interactions with good LY and high transparency in large detectors. Such an approach of extracting the hydrophilic elements (e.g. B, Te, and Li) into organic scintillators cannot be achieved using conventional organometallic technology using carboxylate or phosphate ligands. Water-based detectors could also serve as the near detector for long-baseline neutrino beam monitoring or be used for detection of diffuse neutrino flux from distant past supernovae.

Potential WbLS contributions to proposed LS experiments are

• SNO+ [101]: This experiment plans to use a Te-doped LS to determine the neutrino mass and Majorana or Dirac nature of the neutrino via neutrinoless double beta decay (DBD). Formulation of a high Te-loaded scintillator (3 wt%, phase-II) to increase the sensitivity of a ton-scale $0\nu \beta \beta$ detector using either water-based or diol-complexing loadings,
• LUX-ZEPLIN (LZ) [79]: In the context of the upcoming LZ dark-matter detector a Gd-LS with low U and Th background (<ppt) is developed serving as a veto system for dark matter search. A bench-top sample at 5 wt% is currently under evaluation,
• PROSPECT [80]: Prospect will use a ⁶Li-doped LS (0.1 wt%) for the detection of reactor antineutrinos at short baselines ($\lt 10$ m). The ⁶Li-LS with long stability, reasonable LY and attenuation length, and high background rejection capability via PSD is ready for ton-scale production,
• T2K near detector [81]: A water-based LS with scintillation and Cerenkov features for neutrino beam physics and improved control of neutrino-water cross section uncertainties for neutrino oscillation measurements is tested,
• JSNS2 [82]: A Gd-doped WbLS detector deployed at J-PARC spallation neutron source to search for an eV²-scale sterile neutrino, using both neutrinos and antineutrinos as well as both electron-flavor-appearance and disappearance, and muon-flavor-disappearance channels is under evaluation,
• THEIA [77]: Within the project water-based LS at various wt% loadings are under development for multi-physics topics in nuclear, high-energy, and astrophysics, ranging from a next-generation neutrinoless DBD search capable of covering the inverted hierarchy region of phase space, to supernova neutrino detection, nucleon decay searches, and measurement of the neutrino mass hierarchy and CP violating phase, and
• Medical Imaging: A scintillating water phantom for quality assurance of ion-beam therapy (Patent) or a photon detector (loaded with high Z element) for TOF-PET. The radiation hardness has been studied that a scintillating phantom (5% loading) used for proton therapy treatment plan verification would exhibit a systematic LY reduction of approximately 0.1% after a year of operation, which is highly feasible in such applications [83].

The detector requirements for each experiment are summarized in table 4.

Table 4.  Scintillator detector design for projected experiments. The percentages given for the WbLS correspond to the amount of organic scintillator in the liquid.

Experiment	Detector requirement if using WbLS
SNO+	WbLS (70%+) doped with 3%130Te ($0\nu \beta \beta$ isotope)
LZ	Pure LS doped with 0.1%Gd with U/Th < ppt
PROSPECT	WbLS (80%+) doped with 0.1% 6Li with high PSD
T2K	WbLS (10%)
THEIA	WbLS (1%)
QA phantom for ion-beam therapy	WbLS (1%–5%)
TOF-PET calorimetry	WbLS (10%) loaded with high Z elements

The principal of a mass-producible, cost-effective water-based liquid scintillator using the derivatives of linear-alkly-benzene- or hydrophilic polyethylene oxide-based surfactant has been proven. For the success of WbLS application to large-scale physics experiments, several key characteristics of adequate photon production, long attenuation length, quenching effect and timing property, in addition to chemical stability, need to be demonstrated. The preliminary data show that a 1 wt% WbLS could emit 100 optical photons per MeV with  30 m attenuation length (dominated by scattering) at 450 nm. Nevertheless, on-line purification schemes for the organic-water mixing system need to be developed. A one-ton prototype is currently installed at BNL and studies with calibration radioactive sources, cosmic ray muons, and/or high intensity proton beam (100 MeV–1 GeV) are planned. The goal is to examine the light propagation phenomenon including scattering and quenching effects, direct Cerenkov separation from scintillation and solution stability under various environments. Loading organometallic ions into WbLS to enhance the particle-capture signal is well-demonstrated by SNO+ (Te) and PROSPECT (⁶Li) experiments and ton-scale production of metal-loaded WbLS is under development. Preliminarily loading of gadolinium into the developed WbLS can be achieved at $\gt 0.5$ wt% in bench-top studies; few loading parameters to further improve the scattering length and LY performances still need to be fine-tuned. Although each metal ion favors specific surfactant for loading in scintillator, the principal and procedure are well established.

Already in the early work of Reines and Cowan on first neutrino detectors, metal loaded LS was considered as key technology due to the high number of proton targets and good detection efficiency. Their first experiment searching for neutrinos was located at the Hanford Site [1]. Later the experiment moved to the Savannah River Plant in South Carolina, since the new location provided better shielding against cosmic rays. The Hanford detector of the Reines group used toluene as scintillator solvent basis, terphenyl as primary fluor and α-naphthyl-phenyl-oxazole as secondary WLS. The metal was loaded in the form of cadmium propionate (Cd(C₃H₅0₂)₂). The light transmission of the liquid was about 1 m. The solution deteriorated after about 1 year with the formation of an insoluble crystalline precipitate [43].

Still in Hanford, Reines et al tested alternative components, in particular due to toxicity of toluene and Cd propionate. They searched for a more robust liquid, with optical properties less sensitive to LS handling. As solvent there were tests with mineral oil and triethylbenzene (TEB). With TEB high LYs were observed and attenuation lengths of 10 m were recorded before metal loading. The WLS combination was modified to terphenyl as primary and POPOP as secondary WLS. The carboxylate was also replaced in the scintillator version used for the Savannah River detector. Cadmium octoate (Cd(C₈H₁₅0₂)₂), the cadmium salt of 2-ethylhexanoic acid, had shown reasonable performance. At Cd:H ratios of 0.0015 attenuation lengths of 2 m and more were found with better stability compared to the Cd propionate [43]. This was sufficient for the needs and dimensions of the Savannah River detector at that time.

About 40 years after the first neutrino detection by Reines et al, a slightly different approach was followed by the Bugey 3 experiment [84]. This experiment was designed to measure oscillations of reactor neutrinos using 3 identical detection modules. While most experiments use the gamma signal from Gd after neutron capture, a LS loaded with ⁶Li was used in Bugey 3. An advantage of using Li is the good confinement in space and time of the thermal neutron capture reaction $n+{}^{6}\mathrm{Li}\to \alpha +{}^{3}{\rm{H}}$ with kinetic energy of the α-triton pair of 4.8 MeV allowing efficient background reduction. In addition, there are powerful PSD capabilities, since the alpha and triton are strongly ionizing particles with different time profiles of the scintillation signal compared to electrons or positrons. The detector efficiency in a Li-doped LS can be kept high, even close to the walls of the detector volume.

The Bugey 3 detector had a modular design. Rectangular LS cells (8 × 8 × 85 cm³) were equipped with a photomultiplier at both ends. The combination of 98 cells formed together a detector module containing in total 600 ls of LS. One of those modules was placed at a distance of 15 m from the reactor, two clones at a longer distance of 40 m. The Bugey 3 LS was based on pseudocumene (PC) as solvent doped with 0.15 wt% of ⁶Li. The advantage of using PC, beyond the high LY and associated energy resolution, is the good PSD capability of such a LS. The attenuation length was measured to be $(2.6\pm 0.3)\,{\rm{m}}$ at 440 nm wavelength with a LY of $(31\pm 5)$% of anthracene [85]. Still, the Bugey scintillator showed instabilities, in particular degradation of the attenuation length. The size of the effect was different in the individual cells. The number of photoelectrons observed for a fixed energy deposition dropped by up to 4% per month [84]. Nevertheless, Bugey 3 succeeded to provide high statistics neutrino energy spectra close to nuclear reactors [86].

Few years after Bugey, the CHOOZ [39] experiment tried to measure neutrino oscillations using longer baselines. The reactor neutrino experiments before CHOOZ measured the neutrino flux at distances of about 100 m and below, the CHOOZ experimental site is 1.05 km from the two reactor cores. Therefore a larger detector size was required to obtain reasonable event rates. The target mass of the CHOOZ LS was about 5 tons inside a single acrylic vessel. In such a big LS detector the pathlength of the scintillation light through the detector liquids is increasing, setting stronger requirements on the optical clarity of the liquid.

The CHOOZ LS was doped with 0.9 wt% of Gd in the form of a nitrate. The salt Gd(NO₃)₃ was dissolved in hexanol. This solution was then diluted in a paraffinic liquid (Norpar-15) and isopropylbiphenyl (IPB) with p-terphenyl and bis-MSB as fluors. At the start of the experiment the attenuation length was about 4 m and the LY 35% of anthracene [39]. The scintillator transparency showed significant degradation over time, limiting the livetime of the CHOOZ detector to about one year. Oxidation by the nitrate ion was suspected to be the main reason for the observed LS degradation [39]. Nevertheless, with the data available it was possible to set the most stringent upper limit on the neutrino mixing angle ${\theta }_{13}$ for about one decade.

At the time of CHOOZ another neutrino experiment at a distance of about 1 km from the reactor was constructed at the Palo Verde Nuclear Generating Station, back then the largest nuclear power plant in the US [44]. The antineutrino spectrum was measured using a segmented Gd-loaded LS detector. The detector consisted of a 6 × 11 array of acrylic cells dimensioned at 900 × 13 × 25 cm³ and was filled with a total of 11.3 tons of LS. As with the CHOOZ experiment, Palo Verde did not find any hint for neutrino oscillations and found a similar upper limit on ${\theta }_{13}$ [87].

The final Gd-loaded LS used in the Palo Verde experiment fulfilled the needs of the experiment. The scintillator development however reflected the difficulties and challenges in the production of metal loaded LS. For material compatibility reasons with the acrylic cells, the LS basis PC was diluted with mineral oil (60 vol%). In Palo Verde the metal oxide Gd₂O₃ was converted to the carboxylic acid 2-ethylhexanoate [88] and then loaded to the scintillator at a Gd-concentration of 0.1 wt%.

A first batch of fully blended scintillator did not even remain stable during transport in 200 l steel drums. Therefore, it was decided to mix a PC based concentrate on-site directly with mineral oil and additional PC. The concentrate itself was produced by Bicron company (BC-521C). Special care had to be taken during mixing, since rapid addition of the oil led to irreversible precipitation of the Gd compound. Any disturbance or agitation of the liquid was a risk for the scintillator stability. Even nitrogen bubbling for oxygen removal had to be avoided, not to destabilize the liquid [88].

Whereas the LY was found to be rather robust, the main challenge in the scintillator production was to achieve long attenuation lengths and keep them over long time periods. After several steps of optimization in the scintillator production procedure attenuation lengths of more than 10 m could be achieved. Although some aging effects were still observed, stability of the liquids transparency was acceptable for the needs of the Palo Verde experiment. The LY was measured to be 56% of anthracene [88].

The search for the last unknown mixing angle in the three flavor paradigm of neutrino oscillations triggered the construction of three additional reactor neutrino experiments at the 1 km baseline about 10 years after CHOOZ and Palo Verde. This new generation of experiments has additional near detectors at shorter baselines, which measure the neutrino flux before oscillation takes place. In this way, the systematic uncertainties from flux prediction and neutrino detection can be effectively reduced allowing to measure the oscillation effect. The first indication for disappearance of reactor neutrinos due to a non-zero value of ${\theta }_{13}$ was provided by the Double Chooz experiment [51].

For the Far Detector in Double Chooz, the site of the CHOOZ experiment could be reused. The detector design was improved in several ways. Among the improvements there is the larger target size (about factor 2) and the inclusion of a non-scintillating Buffer liquid as shielding, protecting the sensitive detector volumes from external radioactivity [51]. As solvent for the 10 m³ of target liquid in each detector a mixture of 20 vol% o-PXE and 80 vol% n-dodecane were used [26]. A n-alkane was chosen instead of mineral oil to have a well defined chemical composition. In this way the H ratio can be calculated to a high precision. Furthermore n-dodecane has the advantages that it is highly transparent, the radiopurity had already been proven by the KamLAND experiment, it is compatible with the acrylic vessel and provides a large number of target protons.

In Double Chooz for the first time a metal loaded scintillator based on the BDK chemistry was used on a large scale. The liquid was loaded with 1 g l⁻¹ (0.123 wt%) of Gd which reduces the mean neutron capture time of the undoped liquid by a factor of more than 6 to less than 30 μs. Stable transparency of more than 5 m above 430 nm was observed [26] during several years of data collection [89]. Special care was taken to keep the H₂O level below 50 ppm. The target LY was measured to be 58% of a pure o-PXE based scintillator with the same fluor combination of PPO and bis-MSB [26]. The radiopurity of the liquid was found to be well within specifications.

A similar detector design as in Double Chooz was implemented in the Daya Bay experiment [90]. This experiment provides the most precise measurement of the neutrino mixing angle ${\theta }_{13}$ so far with ${\sin }^{2}(2{\theta }_{13})=0.084\pm 0.005$ [91]. A rather large number of reactors and neutrino detectors are involved. The Daya Bay nuclear power complex in China consists of six cores grouped in three pairs with maximal distances from each other of about 1 km. Neutrinos are measured in three experimental halls, two of them near the reactor cores each housing two identical detectors with 2 × 20 tons of target liquid. In the experimental hall at the far position close to the baseline of maximal oscillation, another four identical detectors are installed. This configuration provides very high statistics compared to other reactor neutrino experiments.

The LS in Daya Bay is based on LAB as solvent and uses the carboxylic acid 3,5,5-trimethylhexanoic acid (TMHA) for the Gd-loading (0.1 wt%) [35]. After thoroughly testing and prototyping of the liquid almost 200 tons of Gd scintillator were produced within 6 months. Solid Gd-TMHA prepared from GdCl₃ was dissolved directly into the LAB. The attenuation length as measured in the storage tanks was about 15 m at 430 nm, a typical wavelength for scintillator emission. The photoelectrons per MeV reduced by about 1.3% per year due to a slight deterioation of the attenuation length [35]. A high LY for a metal loaded LAB based LS of about 50% of anthracene was observed.

The third reactor neutrino experiment measuring the neutrino mixing angle ${\theta }_{13}$ is the RENO experiment [46] in South Korea. As neutrino source six pressurized water reactors are positioned along a 1.3 km long straight line. Two identical antineutrino detectors are located underground at reactor flux weighted distances of 409 m (near position) and 1444 m (far position). As in Double Chooz and Daya Bay a 0.1 wt% Gd-LS was filled into cylindrical acrylic vessels with a volume of 18.6 m³ in each detector.

The composition of the RENO scintillator is very similar to the Daya Bay target liquid. Both use LAB as solvent, PPO (3 g l⁻¹) and bis-MSB as fluor combination and the carboxylic acid TMHA for the Gd loading. A difference between the Daya Bay and the RENO scinitllator production was that RENO used a two phase liquid–liquid extraction technique [92]. GdCl₃ was dissolved in ultrapure water and the pH value of the solution adjusted. Then it was mixed with a neutralized TMHA solution and LAB. The Gd is chemically transferred from the aqueous phase into the organic phase. Finally, the phases need to be separated again. In this way a stable Gd-LS with more than 10 m attenuation length (430 nm) and a primary LY of about 8000 photons/MeV could be produced [92], which corresponds to about 48% of anthracene.

The Nucifer experiment is a very short baseline reactor neutrino experiment running at the Osiris research reactor in France [63]. It was designed in the context of neutrino detectors for safeguard and non-proliferation. Moreover, it fulfills basic requirements to study the reactor neutrino anomaly [93] and sterile neutrinos. With a distance of about 7 m to the compact Osiris reactor core (70 MW thermal power) it can measure about 300 neutrinos/day in reactor ON phases with 850 ls of Gd-LS as Target.

The Gd LS is based on the liquid used in Double Chooz. The same β-diketonate complex Gd(thd)₃ was used to dissolve the metal. Compared to the Double Chooz LS a higher Gd-concentration of 0.17 wt% was used. In this way background can be reduced by shortening the coincidence time window. Background in Nucifer is much higher than in Double Chooz, mainly due to the shallow depth and the reactor vicinity. The higher Gd-concentration, however, increases quenching and lowers the LY. To compensate for this effect, the o-PXE to n-dodecane ratio in the LS was increased to about 50% and more of the primary fluor PPO was added. In this way not only LY, but also the PSD capability are improved.

There are several upcoming reactor neutrino experiments better suited and adapted to search for sterile neutrinos in the MeV range. Many of them as Stereo [94], NEOS [95] or Neutrino-4 [96] will use Gd-loaded LS for the antineutrino detection. Those will profit from the extensive R&D work on Gd-loaded LS from previous experiments. In particular, the successful performance of the Gd-LS of Double Chooz, Daya Bay and RENO are setting a good basis for upcoming scintillator productions. In Stereo and NEOS, DIN is added to the scintillator at concentrations of 5%–10% to increase PSD capabilities. There are also plans to search for sterile neutrinos with ⁶Li doped LS as in PROSPECT [80].

The Borexino history started with first ideas on a LS detector loaded with boron [103]. The goal was to measure the CC and NC interactions of solar neutrinos on ¹¹B nuclei to solve the solar neutrino puzzle. The CC reaction of interest was the inverse beta decay from ¹¹B to ¹¹C. In 26% of the cases the ¹¹C nucleus is left in an excited state. In this way the coincidence of an electron, a gamma from the ¹¹C deexcitation and a final ${\beta }^{+}$-decay of ¹¹C (20 min half-life) is expected. In the NC reaction the signal would be a single gamma from the an excited ¹¹B state. The basic concept of comparing rates measured in CC and NC was similar to the later SNO-experiment. The scintillator technology offered the possibility of a 'live' detector with low energy threshold and good energy resolution. It was reported that a B-loaded scintillator was tested, which contained 80% trimethylborate (10% boron content) and gave a LY of 52% of anthracene with an attenuation length $\gt 10\,{\rm{m}}$ [103].

Finally, due to the extraordinary radiopurity requirements on the LS for solar neutrino detection the Borexino Collaboration changed to a simpler binary LS system with PC as solvent and PPO (1.5 g l⁻¹) as single fluor. To achieve the ultra low background levels, the design of Borexino uses graded shielding, since then it is typical for common large LS neutrino detectors. At the center of concentric shells with increasing radiopurity there are 300 tons of LS [37]. This inner volume is surrounded by another 1000 tons of liquid, a passive shield composed of PC and 5.0 g l⁻¹ DMP (dimethylphthalate), a quencher for scintillation light. More than 2000 eight-inch PMTs are mounted on a stainless steel sphere with 6.85 m radius. The detector design is completed by a water Cerenkov tank as outer veto detector. With this setup, Borexino succeeded in performing the first real-time spectral measurement of sub-MeV solar neutrinos starting with the monoenergetic ⁷Be-neutrinos of 0.86 MeV energy [18]. At a later stage the outstanding radiopurity of the LS even allowed the detection of the low energy pp-neutrinos [104].

The goal of a real-time measurement of low energy solar neutrinos in the sub-MeV region triggered several studies on In-loaded LS over several decades [40–42, 54]. Compared to detection techniques based on the elastic neutrino-electron scattering, In-loaded LS have the advantage of using the inverse beta decay of ¹¹⁵In to the 614 keV excited state of ¹¹⁵Sn as shown in figure 5. The delayed coincidence signature of a prompt electron containing information about the neutrino energy, followed by a delayed (few μs) gamma cascade can be used for distinct background reduction. However, ¹¹⁵In is a ${\beta }^{-}$ emitter with a half-life of $4.4\times {10}^{14}$ a (0.25 Bq per gram of natural indium) and an endpoint energy of about 500 keV, which is a challenge for such detectors. For this reason and due to the additional difficulties of producing a stable and transparent liquid at high indium loading, all early attempts to develop an In-based LS detector were stopped. Nevertheless, the approach of using In-loaded LS for low energy solar neutrino detection was reconsidered in the beginning of this century and got integrated into the LENS activities, which originally proposed different isotopes.

Zoom In Zoom Out Reset image size

The motivation for LENS was a measurement of the low energy pp-neutrino spectrum of the Sun. The majority of the neutrinos in the Sun are produced in this proton–proton fusion reaction. The pp-neutrino flux can be estimated from the luminosity of the Sun and is almost independent of solar model predictions. At the start of LENS three new candidates for low energy solar neutrino spectroscopy were proposed [105]. These were the $\beta \beta$-decay candidates ¹⁷⁶Yb, ¹⁶⁰Gd and ⁸²Se. They all have Q-values low enough to allow for the detection of pp-neutrinos via inverse β-decay. The early stages of the LENS R&D phase mainly focused on ¹⁷⁶Yb (Q-value of 301 keV). The excited state of the daughter nucleus ¹⁷⁶Lu decays under emission of a 72 keV gamma with a half-life of 50 ns. The main challenges for this isotope were the rather short coincidence time and the rather low energy of the delayed gamma. For ⁸²Se the situation in terms of timing and energy is even worse and Gd has other technical difficulties as the background from the alpha emitter ¹⁵²Gd plus complicated level scheme of the daughter nucleus ¹⁶⁰Tb. The main background sources identified for the Yb candidate isotope were ²³⁵U, ¹⁶⁹Yb and ¹⁷⁶Lu. Different Yb-loaded LS based on carboxylic acid systems were produced, but finally also Yb was dropped as target candidate and the collaboration decided to switch to ¹¹⁵In already proposed many years before LENS.

Several groups inside the LENS collaboration were working on In-loaded LS with different carboxylic acids [106, 107] or with β-diketonates [50]. Two of the In-LS were tested along with unloaded reference cells in a low background environment at the LLBF in the Gran Sasso National Laboratory in Italy. The cells contained a carboylic acid version and a β-diketonate. In this setup it was even possible to measure the ¹⁴C content in the scintillator solvent in a cell of only 2 ls volume [108]. For the indium carboxylates good results were obtained with isovaleric (5C) and 2-methyl valeric acid (6C) [106]. The candidate chosen for the LLBF test was 2-methyl valeric acid (MVA) due to its lower toxicity relative to the isovaleric acid. Indium has a high chemical reactivity to –OH groups even at low pH values. Part of the carboxylic –COO-groups are replaced by hydroxyl ions (–OH). Although hydrolysis can finally lead to instabilities and might cause the metal to come out of solution, some fraction of –OH groups helps in terms of LY. The challenge is to keep solubility and avoid strong polymerization in the presence of –OH groups. In fact, in the prototype cell the indium was dissolved in the form of a polymeric hydroxycarboxylate of the form [In(OH)${}_{2.2}$(2MVA)${}_{0.8}$]_n [109]. The –OH fraction can be tuned by pH adjustment in the liquid–liquid extraction. For an indium loading of about 54 g l⁻¹ dissolved in PC at a fluor concentration of 4 g l⁻¹ BPO (2-(4-biphenyl)-5-phenyloxazole) a LY of about 54% of anthracene could be achieved with an attenuation length of 3.9 m at 430 nm. Long term stability of the attenuation length could be confirmed over 6 years. The effective attenuation length of the liquid measured in the 1 m prototype cell was about 5 m [106].

After the LLBF measurement the research on In-loqded LS for LENS continued in the US. Further improvements in terms of the optical properties of a similar indium carboxylate LS based on MVA were reported [107]. After a liquid–liquid extraction at pH 6.88 solid indium carboxylate is produced by vacuum distillation. The solid indium material can be directly dissolved in the scintillator solvent with an indium loading of 8 wt% and above. Even at such high loadings transparencies in the 10 m range at high LYs of 55% were reported after production [107]. Similar to the LS described above, the indium complex contains –OH groups. The molecular formula found for the indium complexes is In(MVA)₂(OH). This composition is consistent with results from independent studies extracting the indium at a similar pH value around 6.9 [57]. After production there are mainly monomers, but after several months dimers, trimers and other oligomers could be identified in the LS. Those can cause a slight decrease of the LS transparency.

The KamLAND was built at the Kamioka Observatory, Japan, an underground neutrino detection facility to detect electron antineutrino oscillation. The KamLAND detector's outer layer consists of a 18 m stainless steel containment vessel equipped with about 2000 17 inch photo-multiplier tubes. Its second, inner layer, consists of a 13 m nylon balloon filled with a LS composed of thousand tonnes of 20 vol% pseudocumene in 80 vol% of n-dodecane loaded with PPO (1.52 g l⁻¹) [111]. KamLAND-Zen is a successor of the KamLAND experiment that used the same detector to study DBD of ¹³⁶Xe from a balloon deployed in the LS in summer 2011.

Xenon does not have long-lived radioactive isotopes other than ¹³⁶Xe. The decay Q-value of the ¹³⁶Xe $\to \,$¹³⁶Ba transition is 2457 keV, among the highest values of $0\nu \beta \beta$ isotopes. Solubility of xenon gas in LS is high, but decreases with increasing temperature. KamLAND-Zen uses 380 kg (3 wt%) of 91.7% enriched ¹³⁶Xe contained in a 1.54 m radius inner balloon, which was contaminated with Fukushima fallout, eventually leading to the leaching of nuclear-decay isotopes, particularly ^110mAg, in its scintillator.

Nevertheless, KamLAND-Zen set a limit for half-life of $1.9\times {10}^{25}$ y and measured the $2\nu \beta \beta$ lifetime as $(2.38\pm 0.02(\mathrm{stat})\pm 0.14(\mathrm{syst})\times {10}^{25})$ y, consistent with EXOs results [112]. KamLAND-Zen plans continued observations with 600–800 kg more enriched ¹³⁶Xe and improved detector components including higher quantum efficiency photomultiplier tubes, light-collection mirror, and new, higher LY LS, LAB+PPO (2 g l⁻¹).

For several reasons the rare Earth element Neodymium (Nd) is a promising canidate for $0\nu \beta \beta$-decay searches. The isotope ¹⁵⁰Nd (5.6% natural abundance) has the shortest calculated half-life of all isotopes for a given effective Majorana neutrino mass [113]. Moreover, the Q-value for ¹⁵⁰Nd is at 3.37 MeV, which is above the energy of typical backgrounds from natural radioactivity. Neodymium is rather inexpensive and can be deployed in large quantities into LS.

Such Nd-loaded LS were for example proposed and prepared in the context of SNO+ [9]. Concentrations of Nd of more than 1% have been tested. Most attempts of Nd loading were based on carboxylic acid versions. For a specific metal loaded LS with a concentration of 6.5 g l⁻¹ Nd dissolved in PC a LY of 75% compared to the unloaded scintillator version was achieved in a small sample and an effective optical attenuation length of 2 m was measured in a 2 l cell [114].

The drawback of Nd are intrinsic absorption peaks in the PMT sensitive region of the spectrum. There is a transparent window at 370–415 nm, where attenuation length can be high ($\gt 5$ m). On the other hand at wavelengths from 415–500 nm several structures can be found in the absorption spectrum of Nd, which can not be removed by purification. Therefore, for a large SNO+ type detector light absorption at ∼1% loadings is too high with common secondary WLSs.

SNO+ is a large LS-based experiment located 2 km underground (5890 ± 94 mw.e.) at SNOLAB, Sudbury, Canada. It reuses the Sudbury Neutrino Observatory detector, consisting of a 12 m diameter acrylic vessel to be filled with 780 tonnes of ultra-pure LS. Designed as a multi-purpose neutrino experiment, the primary goal of SNO+ is a search for the neutrinoless DBD. After initial R&D work on ¹⁵⁰Nd the focus is now on ¹³⁰Te as a target.

The isotope ¹³⁰Te has the highest natural abundance among the DBD isotopes and is of great interest as a $0\nu \beta \beta$ source owing to its low $2\nu \beta \beta$ rate, good optical transparency in the visible region, and high natural abundance (34.1%, $Q=2529$ keV), which negates the need for costly enrichment. Yet it is not feasible to employ Te in LS using the conventional complexing-ligand, –COOH or –P = O, organometallic loading method due to its hydrophilic nature. Several attempts of Te-loaded LS have been made over the past few years, most Te-carboxylates or -diketones dissolved in LSs formed colored solutions, which are not suitable for a large production.

A new loading method using butanediol as complexing ligand for Te extraction at concentration levels of several percent into LAB is currently developed for SNO+. A superior optical transmission with reasonable LY is achieved at 0.5 wt% of Te in LS, which is highly feasible for large scale detector. The diol-complexing method is now selected for SNO+ phase-I deployment. The details of Te-diol preparation will be presented in separate articles.

As mentioned above the water-based LS [76] offers an alternate approach of loading metallic ions in aqueous phase directly into organic scintillator. Telluric acid, Te(OH)₆, is first dissolved in water and then loaded into LAB with the addition of a surfactant or a mixture of two surfactants. The Te-WbLS at 5 wt% has been prepared and stable for 1.5 years with better LY than that of Te-diol-LS. However to escape from color-quenching by self-absorption bands close to 400 nm and reduce the scattering effect induced by addition of surfactants, a secondary WLS added to the scintillator cocktail or an improved purification method of surfactants are required.

In phase I, the detector will be loaded with 0.5 wt% natural tellurium, corresponding to nearly 800 kg of ¹³⁰Te, with an expected effective Majorana neutrino mass sensitivity in the region of 55–133 meV, just above the inverted mass hierarchy. Higher loading development is continuing, the possibility to deploy up to ten times more natural tellurium has been investigated, which would enable SNO+ to achieve sensitivity deep into the parameter space for the inverted neutrino mass hierarchy in the future.

Additionally, SNO+ also aims to measure reactor antineutrino oscillations, low-energy solar neutrinos, geo-neutrinos, to be sensitive to supernova neutrinos, and to search for exotic physics. A first phase with the detector filled with water will begin soon, with the scintillator phase expected to start after a few months of water data taking. The phase I is foreseen to start in 2017.

The impact of background in a $0\nu \beta \beta$ experiment heavily relies on the energy resolution of the detectors. The solid-state or bolometric detectors have exceptional energy resolution that reduces the tail of the $2\nu \beta \beta$ spectrum from overlapping the region of interest of the $0\nu \beta \beta$ spectrum. Moreover, the contamination by for example the radioactive background from the 2.6 MeV γ emanating from ²⁰⁶Tl is reduced. To compensate the deficiency of energy resolution of scintillator detectors (3%–4% at 2.6 MeV), high purity LS and $0\nu \beta \beta$ target are essential for the experiments to succeed.

Very low radioactivity of LSs (U at 10⁻¹⁸ g g⁻¹ and Th at 10⁻¹⁷ g g⁻¹) has been reported by Borexino and KamLAND. KamLAND-Zen will continue to have highly pure ¹³⁶Xe isotopes. Enrichment is achieved with centrifuged method, which moreover frees the material from other inherited gas contaminants. The balloon cleanness is improved as well (U/Th/K at 10⁻¹³ g g⁻¹), in addition to the removal of ^110mAg fallouts from its LS by distillation. For SNO+, tellurium compounds, as an element in the Group-VIA family, typically have relatively low actinide contaminations compared with lanthanide isotopes (e.g. Nd). However, even though Te has low intrinsic U/Th backgrounds, the cosmogenic production of radioactive isotopes, such as sodium (²²Na), cobalt (⁶⁰Co), silver (^110mAg), antimony (^124,126m Sb) and tin (¹²⁶Sn), could still contaminate the $0\nu \beta \beta$ region. Some of these cosmogenic isotopes, i.e. Sn, with high hydrolysis constants can be removed along with U and Th by the self-scavenging method that was previously developed for gadolinium (Gd), neodymium (Nd), and applicable to other lanthanides via their feeding solutions [56]. However, other cosmogenic isotopes such as Co or Ag are very water soluble, which implies that this scavenging method is not effective for their removals. Consequently an acid and thermal recrystallization method for Te compounds prior to its preparation of loading in LS is developed to cleanse these cosmogenic and radiogenic contaminants during the metal-loading process. This purification depletes U and Th and several cosmogenic isotopes from Te feedstock by a factor of ${10}^{2}\mbox{--}{10}^{3}$ in a single pass [115]. The process is also found to improve the optical transmission by removing traces of colored impurities. In addition to the scintillator-based experiments, this cleansing scheme has potential applications to the production of radiopure tellurium crystals for other rare-event experiments.