Neutron/Gamma Discrimination with CLYC and CLLBC

1. The Dual-Mode Detection Problem

Radiation portal monitors, handheld radioisotope identifiers (RIIDs), and physics experiments share a common requirement: distinguishing neutrons from gamma rays in real time, with a single compact sensor. The challenge is that most scintillators respond to both radiation types. A NaI(Tl) crystal, for instance, records a gamma photon and a neutron interaction with nearly identical pulse shapes. Separating the two streams requires either a separate ³He or ⁶Li neutron counter operated alongside the gamma detector, or a dedicated moderator assembly to thermalize fast neutrons before they reach the sensing volume. Both approaches add weight, volume, and complexity.

Elpasolite crystals, specifically CLYC (Cs₂LiYCl₆:Ce) and CLLBC (Cs₂LiLaBr_4.8Cl_1.2:Ce), solve this at the material level. A single crystal simultaneously captures thermal neutrons and gamma photons, and produces scintillation pulses whose time profiles are distinct enough that digital signal processing can assign each event to one category or the other. The result is a compact, single-readout detector capable of simultaneous gamma spectroscopy and neutron detection, no secondary counter required.

This paper explains the physical mechanism behind that discrimination, defines the figure of merit (FOM) used to quantify its quality, and describes the specific properties of CLYC and CLLBC that make them the preferred choice when dual-mode performance matters.

2. How Pulse-Shape Discrimination Works

2.1 The Scintillation Pulse

When an ionizing particle deposits energy in a scintillator, it excites the crystal lattice and the luminescent centers (usually Ce³⁺ or Tl⁺) emit photons over a characteristic time profile. That profile is not a single exponential. In practice, it is a sum of components with different decay constants, each weighted according to how the excited state was populated. Gamma rays interact primarily via Compton scattering and photoelectric absorption, creating fast electrons. Those electrons deposit their energy in a way that preferentially populates fast-decaying singlet states. Neutrons, by contrast, produce recoil protons (fast neutrons) or lithium and alpha fragments (thermal neutrons via the ⁶Li(n,t)α reaction), and heavier charged particles are more densely ionizing. Dense ionization populates slower triplet and trap states at a higher rate. The tail of the pulse is therefore longer and brighter for neutron events than for gamma events at the same total deposited energy.

PSD exploits this difference. Two quantities are measured per pulse: the total integrated charge (Q_total), which is proportional to the deposited energy, and the charge in a late time window (Q_tail), which captures the slow component. The ratio Q_tail/Q_total, sometimes called the tail-fraction or PSD parameter, separates neutron events from gamma events when plotted against Q_total. The result is a two-dimensional scatter plot with two distinct populations.

2.2 The Figure of Merit

Quantifying how well separated those two populations are requires a single number. The standard definition is:

where S is the separation between the centroids of the neutron and gamma bands in PSD-parameter space, and FWHM_n and FWHM_γ are the full-widths at half-maximum of each band. A FOM above 1.0 indicates that the two populations are resolved with less than roughly 0.1% overlap at their base. FOM values above 1.5 to 2.0 are generally considered excellent for practical instruments; misclassification rates fall into the parts-per-thousand range or better. Below FOM 0.8, discrimination becomes unreliable and coincidence methods or additional moderator shielding may be needed.

Several factors influence the FOM a given crystal achieves in practice. Light yield matters because higher photon output reduces photon-counting statistics, sharpening each band. The ratio of fast-to-slow decay components must differ enough between particle types; if the crystal has only one scintillation pathway, there is nothing to discriminate. Detector noise, photomultiplier transit-time spread, and digitizer sampling rate all broaden the bands and degrade the FOM. Crystal purity and the uniformity of Ce³⁺ activation also affect how consistently a given particle type produces the same pulse shape.

3. The Elpasolite Advantage

3.1 Crystal Structure and Neutron Sensitivity

CLYC and CLLBC belong to the elpasolite family, characterized by a double-perovskite structure with the general formula A₂BB'X₆. In both materials, lithium occupies the B' site at high concentration. Lithium-6 has a thermal neutron capture cross section of approximately 940 barns via the reaction:

Both the alpha and triton fragments are densely ionizing. They deposit their combined 4.78 MeV energy over very short tracks inside the crystal, producing a high local ionization density. As described in Section 2.1, this dense excitation strongly populates the slow scintillation component, giving neutron events a distinctly longer tail. Gamma-induced Compton electrons are far less dense in ionization and produce a faster pulse. The contrast between the two regimes is large enough in elpasolites that FOM values well above 1.0 are achievable with standard photomultiplier readout and modest digital filtering.

CLYC achieves this with a composition of Cs₂LiYCl₆:Ce, where yttrium provides the crystal host lattice and chloride anions enable reasonable transparency at the Ce³⁺ emission wavelength near 370 nm. The density is 3.31 g/cm³, the refractive index is 1.81, and the decay structure features three distinct components at roughly 1 ns, 50 ns, and 1,000 ns. The relative light yield is 30 to 40 (on the NaI(Tl) = 100 scale), which is modest compared to NaI(Tl) but sufficient for clear PSD at gamma energies above roughly 200 keVee. Hygroscopicity requires hermetic encapsulation, though standard aluminum or stainless housings are well established for this crystal family.

CLLBC extends the elpasolite concept by replacing yttrium with lanthanum and partially substituting bromine for chlorine: Cs₂LiLaBr_4.8Cl_1.2:Ce. The result is a significantly brighter material (relative light yield 70 on the same scale) with better energy resolution. The density rises to 4.08 g/cm³, the emission shifts to about 420 nm, and the dominant decay components fall around 120 ns and 500 ns. The higher light yield sharpens both the PSD bands and the gamma energy resolution, making CLLBC particularly attractive when the instrument must perform simultaneous isotope identification and neutron detection at good resolution.

3.2 Three Scintillation Channels in One Crystal

What makes elpasolites unusual is that a single crystal volume supports three distinct detection modes. First, thermal neutrons are captured via the ⁶Li reaction and produce the high-PSD-parameter events already described. Second, gamma rays scatter and photoabsorb, producing the low-PSD-parameter population. Third, fast neutrons above roughly 0.5 MeV interact with hydrogen-containing materials in the housing or via direct recoil in the crystal chlorine and cesium nuclei; these events appear at intermediate PSD values and require either moderation or coincidence gating to extract cleanly. In most fielded RIID instruments, a thin polyethylene moderator disk (a few centimeters) thermalizes fast neutrons before they reach the crystal, so the detector effectively sees a clean thermal neutron beam and a gamma field simultaneously.

This three-mode sensitivity, governed by a single photomultiplier or SiPM readout, is the fundamental reason elpasolites have displaced ³He tubes in next-generation handheld identifiers. There is no high-pressure gas to handle, no separate preamplifier chain for the neutron counter, and no alignment requirement between two detector heads.

3.3 Comparing CLYC and CLLBC for Practical Use

Note: Energy resolution values for CLLBC are based on general literature for this crystal class; confirm against specific crystal lot and readout configuration before citing in instrument specifications.

4. Digital PSD Implementation

4.1 Charge Integration Windows

Modern multichannel analyzers such as the Berkeley Nucleonics bMCA implement PSD in firmware using programmable charge integration windows. When a pulse arrives, the digitizer captures the full waveform at a sampling rate typically between 125 MHz and 500 MHz. A short integration gate (commonly 50 to 80 ns wide) captures the prompt component of the pulse; a longer gate extending 500 ns to several microseconds captures the total charge including the slow tail. The ratio of late-window charge to total charge is computed per event and stored alongside the energy value, forming the two-dimensional scatter histogram shown in Figure 1.

Gate timing is the primary tunable parameter. If the short gate is too narrow it clips the gamma peak and reduces separation. If the long gate extends unnecessarily, dark current noise and pile-up at high count rates degrade both bands. For CLYC, where the slow component extends to 1 microsecond, a 1.2 microsecond total gate with a 60 ns prompt gate is a commonly cited starting point; for CLLBC with its shorter dominant components, a tighter total gate often delivers adequate separation with less sensitivity to pile-up. Both materials benefit from validated gate settings at the specific count rate and detector geometry of the application.

4.2 Applying the PSD Cut

Once the scatter plot is populated, a threshold is set in PSD-parameter space (the dashed red line in Figure 1). Events above the cut are classified as neutrons; those below are gamma. The threshold position involves a tradeoff: moving it upward reduces gamma misclassification into the neutron count but lowers neutron detection efficiency. In regulatory contexts such as ANSI N42.34 or IEC 62327, specific sensitivity and selectivity requirements constrain where the cut can sit. Instrument firmware typically offers a user-adjustable sensitivity setting that moves this threshold, with the default factory setting targeting the standard's minimum neutron sensitivity.

For high-count-rate environments (portal monitors processing tens of thousands of events per second), digital pile-up rejection is essential before the PSD algorithm runs. A pulse that represents two overlapping events will produce a spurious PSD value and appear in a region of the scatter plot between the two bands. Pile-up rejection circuitry, or software that flags pulses with anomalous rise times, keeps these events from polluting the classification.

5. System-Level Design Considerations

5.1 Readout Selection: PMT or SiPM

Both photomultiplier tubes and silicon photomultipliers work with CLYC and CLLBC, but the choice affects PSD performance. PMTs with bialkali photocathodes have good quantum efficiency near the 370 nm emission of CLYC and good single-photon timing resolution. Transit-time spread in the 0.5 to 1.5 ns range is the main PMT contribution to pulse-shape blurring. For CLYC, where the short (1 ns) component is so fast that it is near the limit of standard PMT timing, that transit-time spread has some impact but the dominant PSD contrast comes from the 50 ns and 1,000 ns components, which are far more forgiving.

SiPMs offer compact geometry, low supply voltage, and ruggedness against magnetic fields. The primary concern for PSD applications is SiPM dark count rate: at room temperature, dark pulses occur at hundreds of kilocounts per second per square millimeter, and they pile onto the slow tail, increasing apparent Q_tail for events occurring shortly after a dark pulse. Careful threshold setting and coincidence timing can mitigate this, and at low temperatures dark count rate drops rapidly. CLLBC, with its higher light yield and longer dominant decay components (120 and 500 ns), is generally more tolerant of SiPM dark noise than CLYC in the same geometry.

5.2 Crystal Volume and Count Rate

Larger crystals improve detection efficiency but increase the probability of multiple simultaneous interactions and make pulse pile-up more frequent at high gamma fluxes. RIID applications typically use 1-inch to 2-inch diameter crystals. Portal monitors may use arrays of 2-inch or 3-inch cylinders. At count rates above roughly 50,000 cps in a 2-inch CLYC crystal, pile-up begins to smear the PSD scatter plot and FOM degrades measurably. Digital pile-up rejection at the MCA level (as implemented in the bMCA) or external shaping cuts the effective duty cycle in exchange for cleaner separation at high flux.

5.3 Moderator Geometry for Fast-Neutron Applications

The ⁶Li(n,t)α reaction has its peak cross section for thermal neutrons. In most real-world neutron threat scenarios (special nuclear materials in shielded configurations, illicit trafficking), neutrons arriving at the detector have energy distributions extending from thermal up to several MeV. A few centimeters of polyethylene in front of the crystal converts a significant fraction of fast neutrons to thermal. The moderator thickness involves a tradeoff: very thick moderators thermalize more fast neutrons but attenuate the gamma flux of interest (high-energy gammas from SNM) and increase size and weight. A common compromise for handheld instruments is 1 to 2 cm of polyethylene.

CLLBC's higher density (4.08 vs. 3.31 g/cm³ for CLYC) also gives it marginally higher gamma stopping power per unit volume, an advantage when the same crystal must deliver good gamma spectroscopy efficiency alongside neutron detection.

6. Performance in Context

6.1 How Elpasolites Compare to Alternative Approaches

Before CLYC became commercially available, the standard dual-mode approach paired a NaI(Tl) or LaBr₃(Ce) gamma detector with a separate ³He proportional tube or a ⁶Li-doped glass scintillator. Helium-3 tubes offer essentially zero gamma sensitivity (with appropriate discrimination electronics) and very high intrinsic neutron detection efficiency, but supply constraints, size, and the complexity of a two-detector module drove the industry to pursue single-crystal alternatives. ⁶Li-glass, listed in the ScintIQ materials portfolio, is non-hygroscopic and fast (60 ns decay), but its relative light yield of 4 to 6 severely limits gamma spectroscopy quality. It is well-suited to purely neutron-counting applications where gamma rejection (rather than coincident gamma spectroscopy) is the goal.

CLYC and CLLBC uniquely occupy the space where good gamma spectroscopy and reliable PSD coexist in a single volume. LaBr₃(Ce) offers better intrinsic gamma resolution (around 2.7% at 662 keV) than either elpasolite, but it has no useful neutron discrimination capability and contains La-138 (a natural radioactive isotope) that adds a background line to its own spectrum. CLLBC substitutes a fraction of the La³⁺ with Li⁺ and mixed halide, gaining neutron sensitivity while preserving much (though not all) of the high light yield that makes LaBr₃(Ce) attractive for resolution. The resulting gamma resolution of CLLBC is better than CLYC and better than standard NaI(Tl), placing it between NaI(Tl) and LaBr₃(Ce) in resolution but ahead of both in dual-mode capability.

6.2 Regulatory and Standards Context

Instruments destined for homeland security, border protection, and nuclear safeguards applications must typically satisfy ANSI N42.34 (handheld RIIDs) or ANSI N42.35 (portal monitors). These standards define minimum neutron sensitivity in counts per second per unit neutron emission rate from a Cf-252 source at a specified distance, alongside alarm false-positive rates at realistic gamma background levels. The minimum neutron sensitivity requirements in N42.34 are achievable with CLYC crystals in the 1-inch to 2-inch range when paired with a polyethylene moderator and a properly set PSD algorithm. CLLBC, with its higher light yield, can meet the same requirements with somewhat tighter FOM margins, providing additional headroom for maintaining performance as the crystal ages or at elevated temperatures. Specific sensitivity numbers depend on detector configuration and must be validated against the published standard for each instrument design.

7. ScintIQ Configuration for PSD Applications

Berkeley Nucleonics offers CLYC and CLLBC as part of the ScintIQ custom detector program, configured with hermetic metal housings, ruggedized optical coupling, and matched readout electronics. Standard configurations pair the crystal with a PMT and voltage divider assembly, with SiPM-coupled variants available for applications where magnetic immunity or compact geometry is required. The bMCA Ethernet-connected multichannel analyzer includes hardware-accelerated PSD in its firmware, with user-programmable gate widths, threshold settings, and real-time two-dimensional histogram generation.

Key considerations when configuring a system for PSD work include crystal size (larger volumes improve efficiency but require careful pile-up management), enrichment level of the lithium (CLYC and CLLBC produced with 95% or higher ⁶Li enrichment are standard for neutron applications), and housing window material (quartz UV windows are preferred over standard borosilicate for optimal transmission at the 370 nm CLYC emission peak). Hermetic epoxy-sealed assemblies are also available for lower-cost, non-field-service applications.

For applications that require both the best achievable gamma resolution and reliable dual-mode discrimination, CLLBC is the recommended starting point. For applications where cost and established supply chain take priority over marginal resolution improvement, CLYC remains the industry-standard choice with a longer track record in fielded RIID instruments.

8. Summary

Pulse-shape discrimination in CLYC and CLLBC works because thermal neutrons captured via the ⁶Li(n,t)α reaction produce densely ionizing alpha and triton fragments that preferentially populate slow scintillation components. Gamma-induced Compton electrons are less dense, populating faster components. Integrating the charge in a prompt versus a delayed time window, then taking the ratio, separates the two event classes in a two-dimensional scatter plot. The figure of merit, defined as centroid separation divided by the sum of band widths, reaches values above 1.5 in well-configured elpasolite systems, placing misclassification rates well below practical concern.

CLYC provides a proven, cost-effective path to dual-mode detection with established RIID heritage. CLLBC extends that capability with higher light yield and better gamma resolution, making it the preferred material when simultaneous isotope identification and neutron detection at high spectroscopic quality are both required. ScintIQ configurations from Berkeley Nucleonics cover both materials with matched readout electronics, hermetic packaging, and support for integration into ANSI-compliant instrument designs.