The purpose of this page is to describe the unique luminescence spectroscopy system at the University of St Andrews. The system is know as the RLTLCL system, which stands for RaioLuminescence, ThermoLuminescence, CathodoLuminescence (RLTLCL). Its detectors have been developed in a partnership with the manufacturer (Photek Ltd) to provide state-of-the-art photon sensitivity combined with fast gating of the power to allow time-resolved spectroscopies. The instrument measures several types of luminescence spectroscopy, including simultaneous multiple excitation, in both continuous wave and pulsed excitation modes as a function of temperature from 20-673 K. Finch et al 2019 RLTLCL system is a published description of the system in the journal Luminescence published in 2019, but the data below include further specific details about how the system was used and its construction that supplement the published article. A detailed description of the detectors is given, and also B-W versions of the diagrams used in the Finch et al. (2019) paper for those with colour blindness. Finally the data formats generated by the system are explained.
1. Introduction and History
In the early 1990s a system dedicated to the measurement of TL was designed with detection of wavelength multiplexed luminescence made possible by the use of photon imaging detectors and housed at the University of Sussex, UK. The high f number optics (f/2.2) allowed the use of low heating rates for the TL so that there were minimal uncertainties between the heating stage temperature and the emission region at the surface of the samples. This was highly effective and, in addition to recording spectrally resolved TL over the wide temperature range, there were unexpected bonuses in peak temperature shifts related to pairing and clustering of impurities, or their association with different defect species. Since different polymorphs inevitably show different emission spectra, it was possible to detect phase transitions very rapidly in a dynamic heating or cooling run. Not only did this reveal hysteresis, but it showed a wide range of previously unsuspected crystalline transformations in materials as diverse as KTP, SrTiO3, fullerenes and superconductors. CL was used to probe the role of contaminants and surface relaxations. Dramatic changes in intensity were noted not only for phase transitions of the host materials, but also driven by phase transitions within inclusions of nanoparticles of impurities (or added dopants). This emphasises that there are very long range interactions from such inclusions that modify the entire sample.
On the retirement of Townsend, the system was brought to St Andrews, where new detector systems were installed, several improvements were made and new software was written. There is now a need to describe the modified system with its enhanced sensitivity and greater capabilities. Rather than describe the history of modifications to the system, we provide here a description in its current state, without differentiating between the original design and our subsequent modifications.
2. The Sample Chamber, Light Path and Stages
The system comprises a central chamber made out of aluminium (typically 3 cm thick) and brass (mm thick) to reduce x-ray leakage (Fig. 1). A USB webcam inside the chamber allows visualisation of the sample on the stage. A two-stage vacuum system roughs down to 10-1 mbar (10 Pa) in a few minutes with an Edwards Rotary Pump and harder vacuum down to 10-6 mbar (0.1 mPa) after a few hours using a Edwards 100/300 650 W Diffstak diffusion pump. The vacuum status is monitored using Edwards VSK1B vacuum switches and two Edwards APG100 active pirani gauge heads. The vacuum switches close when the vacuum is below a threshold (typically 10 mbar = 1 kPa) and the active vacuum heads provide output voltages between 0-6 V as an exponential function of vacuum. These voltages are monitored using a Velleman K8055 USB Experimental Interface board (http://www.velleman.eu/products/view/?country=be&lang=en&id=351346) and converted to an estimate of vacuum – the digital input channels are connected to the vacuum switches and other microswitches around the instrument.
The two-stage vacuum system is controlled by switching mains power on and off to the pumps (rotary and diffusion pumps) using a Measurement Computing USB-ERB08 Electromechanical Relay Interface (http://www.mccdaq.com/usb-data-acquisition/USB-ERB08.aspx). The ERB08 unit also operates the helium compressor, venting during sample change and the safety interlock on the x-ray generator. The stage mounted is detected using microswitches and the program then automatically communicates with the appropriate stage. The system is interlocked for radiation safety and protection of the detectors.
To cover the entire spectrum from the UV to the IR, two detectors operate simultaneously, targeting the UV-Blue and visible-near IR regions, known as the ‘blue’ and ‘red’ detectors respectively. Each is housed in separate detector housings, either side of the chamber (Fig. 1). The light from the sample is focussed by two sets of two fused silica (‘Spectrosil B’) plano-convex lenses, a combination which provides 1:1 imaging of the sample on the entrance slits of the spectrometers. The lenses seal directly to the sample chamber, eliminating the need for additional windows in the sides of the chamber. In the case of the ‘UV-blue’ detector (see below), fused silica elements are necessary to enable UV transmission, but a further advantage of this material is its good resistance to radiation darkening. Since the lenses receive a considerable dose over time from scattered radiation in the chamber, fused silica is used for both lens combinations, even in the ‘red’ detector where UV is not measured. Lens combinations, rather than single lenses, are necessary to match the high f number of the spectrometers (f/2.2). Between the lenses and the entrance slits are lightproof boxes that accept standard (50 mm x 50 mm) filters. The entrance slits to the spectrometers can be swapped to modify the signal strength and spectral resolution, but the typical analysis takes place with 500 mm slits providing a typical wavelength precision of ~2 nm but widths of 125, 250 and 1000 mm are available. Manual shutters are placed in the light paths after the filter boxes.
The focal position of the spectrum is a function of wavelength, and hence the distances from the sample to each detector are different, optimised for the central wavelength region that each detector analyses. The light is diffracted by American Holographic Chemspec 100S gratings of the Rat-field holographic type; one is used nominally for the range 200-450 nm (ref: 446.02/L) with a dispersion of 8 nm/mm and the other (ref: 446.14/L) from 400-800 nm with a dispersion of 12 nm/mm. The numerical aperture is f/2.2. The peak efficiency of each grating occurs at ~350 nm and ~730 nm, chosen near the centre of each spectral range, biased to enhance signals where the red photocathode sensitivity is falling steeply.
Second-order scattering of lower wavelengths into the first order spectrum will occur for the red grating detector, therefore a Precision Optical Instruments GG400 400 nm long pass filter is permanently inserted into the light path of the red detector to remove the short wavelength signals (Fig. 1). The filter is not entirely opaque between 397-400 nm and for samples with a strong signal in this region, we observe a small second order response between 795-800 nm. Rather than attempt difficult corrections for this artefact, we accept that this remains in processed spectra of samples that are strong in the UV-Blue. The two detectors provide two separate sets of spectral data with a substantial overlap from which subsequent data processing creates a single composite UV-IR spectrum.
The chamber is furnished with two stages for cryogenic (20-300 K) and high-temperature (300-673 K) measurement. A third stage operating from -40 to +400oC is also available for samples which have TL transitions around room temperature. The cryostage enters from below the chamber and the high temperature stage at the front; each has a blanking plate to isolate the chamber when the other stage is in use. The stage is a Cryophysics M22 cryostage, cooled by a CTI Cryogenics 8200 water-cooled Helium Compressor. Temperature is controlled by a Eurotherm 2404 controller connected to a Au-Fe thermocouple. The maximum ramp rate is 0.1 K s-1. The cryostage is operated only when the sample has been in vacuum (<10-5 mbar = <1 mPa) for some hours, to avoid frosting of the sample. The higher temperature stages have a Nichrome strip 12 x 50 x 0.7 mm and a chromel/alumel thermocouple controlled by a Eurotherm 818P controller with switching accomplished by a Eurotherm thyristor unit model 462. The maximum ramp rate of the high temperature stage is 3 K s-1.
3. The Detectors
The diffraction gratings deliver a strip of light onto the detector, such that the position of the photon arrival is a function of wavelength. The detectors are imaging plate detectors (ipds) manufactured to our specifications in a collaboration with Photek Ltd (www.photek.co.uk). Ipds were chosen to provide state-of-the-art sensitivity whilst also allowing sub-ms switching of the power to the detector plate for time-resolved luminescence experiments (see below). Cooled detectors were not favoured due to avoid frosting of the detector – the difficulties associated with keeping the detector under vacuum or dry nitrogen were considered to overwhelm the benefits of lower dark current. IPDs use standard PM tube photocathode materials; however electron multiplication is provided by microchannel plates (MCP) rather than a dynode chain. Charge clouds emerging from the channels are proximity focused onto a resistive anode. The current flow resulting from the original photon event is measured at four points on the resistive anode, enabling the position of the photon arrival and its arrival time to be reconstructed. Each ipd has a separate power unit and communicates to the master computer via USB. The ‘blue’ detector is a bialkali photocathode ipd; the ‘red’ detector is a S25 based photocathode – each detector type was chosen to maximise signal and minimise noise in the relevant spectral range. The dark current on each detector was typically 2.4 x 10-4 (blue) and 1.6 x 10-2 Hz nm-1 (red) but we observe a halo effect such that the dark current is lower in the centre and enhanced towards the edges. Following discussions between St Andrews and Photek, this foreshortening of the image was ameliorated by an optional software patch. A fuller description of the imaging plate detectors is given in Supplementary Information 1. The position on the ipd is calibrated against wavelength using Hg, Kr emission lines and laser pointer sources and the output from the detectors are both raw images and reconstructed wavelength. Black body radiation in the infrared is seen as a background in all runs using the high-temperature stage above ~600 K and becoming progressively more intense as temperature rises to the maximum (673 K).
The system and detectors used here exploit the high sensitivity and dynamic range of photocathode detectors. However, they have a weakness in that their performance falls at long wavelengths. By contrast one can use CCD signal collection which functions into the near infrared region. Our system has sufficient access ports that in principle it is simple to add a fibre optic link to a longer wavelength CCD spectrometer for those materials which require such data.
4. Sample Excitation
Three forms of excitation are available; 1) a Philips MCN-101 ceramic x-ray tube placed at the rear of the chamber, 2) an electron gun (taken from a Jeol electron microscope) on the top and 3) LEDs inserted inside the chamber and controlled by power sources externally. The x-ray and electron gun sources were controlled manually by units in the electronic rack. The x-ray controller is not operated above 30 kV and 15 mA to avoid x-ray leakage; typical operating conditions are 20 kV and 4 mA, providing dose rates of 1.8 Gy min-1. The electron gun operates between 10 and 25 kV acceleration voltage with a typical beam current of 200 nA and a spot size typically with a diameter of 3 mm, providing an incident power density of ~1 kW m-2. The beam is focussed using the electron optics inherited from the electron microscope source, controlled by external power supplies. Beam position (x-y) is tweaked using external magnetic fields generated by solenoids controlled by power supplies; beam pulsing is achieved by sudden increases in the current applied to these, flipping the beam off the sample. The LEDs are sourced either from RadioSpares (uk.rs-online.com) or Roithner LaserTechnik GmbH (www.roithner-laser.com) and controlled via an amplifier built in house and operated by a logic signal to be either on or off. Emission wavelengths down to 300 nm have been used although deeper UV LEDs are now available. Continuous wave photoluminescence (PL) is difficult since the excitation would saturate the sensitive detector plates. It can be achieved with the use of notch or long pass filters, or by exciting in the UV and only using data from the Visible-NIR detector. Time-resolved PL is achieved without the use of filters by <ms switching of the LEDs coupled to rapid switching of the detectors.
5. System Software
Software was written in house using LabVIEW 2016 (www.ni.com/labview/) to operate from a Graphical User Interface (GUI). All automated components of the system are controlled by LabVIEW drivers but some were provided in other languages (e.g. visual basic or C) and adapted using LabVIEW wrappers. LabVIEW drivers for the stages were downloaded from National Instruments and adapted for the present system. The Eurotherm 818 controller (high temperature stage) was controlled using a version of an older 808 driver by enabling extra commands compatible with the 818. The Eurotherm 2404 controller (low temperature stage) was successfully operated via a generic 2400 series driver. Both controllers work via RS232 protocol, operated via USB using a 4-way USB-RS232 converter.
The detectors came with Photek software (Image32) which can be operated directly or controlled from LabVIEW using wrappers provided by Photek. Photon arrivals were accumulated over preset integration periods, assigning a position on the imaging plate and a time stamp to every photon arrival. The output is a 16-bit monochrome image covering the entire detector area. Photon arrival events outside the rectangular area illuminated by the grating are rejected. Direct image transfer from the Photek software to LabVIEW results in the loss of the most significant bit in the data, and hence the program writes the image data to the hard drive from Image32 and reads it back into LabVIEW as a 16-bit image using routines in the LabVIEW Image Module. For continuous wave (CW) measurements, the time stamps are ignored and the two-dimensional image from each detector is converted to a 1D spectrum by summing individual pixel values onto the wavelength axis.
For time resolved (TR) analysis, excitation is pulsed and photon arrivals in the ‘off’ cycle are analysed as a function of wavelength and arrival time. Since the power to the ipd is switched off during this cycle, time-resolved photoluminescence (TR-PL) is possible without any filters since the intense light from the LED arrays do not encounter the detector whilst the voltage is on, thereby avoiding damage to the detector. TR analysis uses LED (time-resolved PL, TR-PL) or electron beam excitation (TR-CL). The electron beam is pulsed by flicking the beam off the sample using an electromagnet on the column; LEDs are controlled by pulsing the power to them. We observe ~1 ms phosphorescence from standard LEDs after the power is stopped. Although nano-LEDs with rapid switching are available, we have found it more convenient to use cheaper standard LEDs and insert a software delay (typically 3 ms) between the end of the ‘on’ excitation signal and powering up the detectors. Blank measurements with the delay show no sign of the primary excitation in TR-PL. Photon arrival times during data acquisition are binned over multiple iterations producing a 16-bit rectangular image with wavelength on one axis and numbers of photons per time bin on the other. Time bin widths are software controlled but are typically 250 ns, 200 bins and integration times of 5 minutes provide data that we have processed successfully (see below) to provide estimates of lifetime in the ms to s range.
Longer (seconds to hour length) lifetimes are measured using a ‘phosphorescence’ routine in LabVIEW whereby repeated CW spectra are collected and then amalgamated to create a 2D image. The outputs from this type of analysis emulate those of TR analysis so that the same software routines are applied.
6. Offline Data Processing and Manipulation
It is our policy to store data as raw files and system corrections are performed off line. This allows alternative system corrections and data manipulation to be performed retrospectively without the need to reconstruct raw images. We have preferred this approach particularly since offline processing software is regularly improved. The data from the individual detectors are corrected first by subtracting the dark current (as a function of wavelength and exposure time), then dividing by the system responses of the two ipds (Fig. 2, provided by Photek). The gratings are blazed for optimum efficiency in the centre of the range and fall to about 50% at half and double the blaze angle. We model the grating efficiencies as a parabolic function of log of wavelength centred on the blaze wavelength. The efficiency across the majority of the data range is >90%, and might reasonably be considered essentially constant. However there is a significant dip below 250 nm, and an inconsistency between the efficiencies of the two gratings in the overlap region. We therefore correct the data using our grating efficiency model. Although this is a first order correction, it amplifies the deeper UV data and improves the fitting of the two spectra (R2 are increased, see below). The effects of the transmission of the quartz lenses (and the 400 nm long pass filter in the red detector) are also accommodated. The two spectra are matched in the overlapping spectral region, typically 420-480 nm, avoiding the edge regions of both spectrometers where the dark current is high and image foreshortening is observed. Because of the steric effects of the sample and different focus distances to the two imaging plates, the amount of light arriving at the two detectors is never perfectly matched in intensity. We scale the two images with respect to each other to accommodate, with the overlap region comprising a linear combination of the two images. We calculate a fit index (R2) to express how the two images match with linear scaling. Where significant signal is present, R2 values >95% are observed. Analysis of single emission band profiles in energy space give Gaussian fits, suggesting that our system corrections provide accurate and undistorted final profiles. At temperatures above 600 K (~320°C), significant black body radiation is observed in the spectra at 800 nm which increases and creeps to shorter wavelengths with increasing T. The black body radiation from the stage is often many orders of magnitude greater that the luminescence signals from the samples, hence the data for the high temperature stage are deleted where significant black body radiation occurs in a sample blank.
The final outputs are transferred automatically to ORIGIN for plotting and standard 3D and contour plots (used in the present study) are generated automatically. Importing RLTLCL data into Origin 11 May 2015. These are typically contoured with ‘hot’ colours to express higher counts and ‘cold’ colours as lower ones (e.g. Figures 3, 4). However in recognition of the many in the science community who are colour blind (including one of us), the software also generates monochrome contour-based outputs. Examples of the raw and processed data formats are here.