1. INTRODUCTION

Gemologists are looking for accurate, fast, and simple methods to identify gemstone materials. Unfortunately, technical advances in gemstone synthesis and treatments have increased the complexity of their identification. One of the main concerns for consumers is the undisclosed replacement of natural diamonds with inexpensive simulants and laboratory grown (synthetic) diamonds [1–3]. Furthermore, the color of diamonds can be modified using different treatments [4–11]. Treated and laboratory grown diamonds are legitimate gem products as long as they are clearly identified as such to the consumer during sale. However, their similar appearance and lower value relative to analogous natural diamonds can lead to deceitful cases where they are instead classified as natural. Their separation from natural goods has become a pressing challenge for the ${\sim}80$ billion US dollars retail diamond industry, leading to demand for both gemological laboratory services and simple and accurate screening instruments that can be used by members of the trade [3].

The largest group of gemstones in the jewelry market are “melee” diamonds, which weigh less than 0.20 carats [1 carat (ct) = 200 mg] and account for approximately 90% in volume, yet only 5% in value of the market [12]. Despite the relatively low individual value of melee stones, the discovery of an undisclosed laboratory grown or treated diamond in a jewelry piece could have a significant detrimental impact on a jewelry brand’s reputation. Melee diamonds are considered to be at a higher risk of undetected mixing as their small sizes make them more affordable to synthesize and less likely to be tested.

Although the natural diamond gem trade is dominated by colorless to light yellow or brown color diamonds (graded ${\rm D}$ to ${\rm Z}$), pink colors are increasingly popular due to their appealing appearance and rarity [13–15]. Pink colored diamonds can be created through multi-treatment (including irradiation and annealing) of certain diamonds, and these goods are easily found in the trade [1,2,4,6,8,10,15–18]. Due to their high value and popularity, screening equipment to separate natural untreated from multi-treated natural or laboratory grown pink diamonds is also of interest.

Colored gemstones are non-diamond gem materials, which show more than 2000 different hues. Since certain minerals may exist in several different colors, which may overlap with that of other species, their visual appearance is often insufficiently distinctive for their conclusive identification [19]. Historically, this has led to well known cases of famous stones being mistakenly associated with incorrect minerals. For example, many prized “rubies” decorating European crown jewels, such as the Black Prince’s Ruby and the Timur Ruby, are actually much less valuable spinels.

Gemological laboratories are playing a key role in maintaining transparency in the gem industry by analysis through their laboratory identification services, in conjunction with producing identification or screening devices that can be used by the trade. “Identification” refers to conclusively determining whether a material is, for example, natural, treated, laboratory grown, or a simulant, whereas “screening” refers to being able identify one or more subsets of the materials, while referring for further testing those that could not be identified and are at risk of belonging to the other groups. Traditionally, gemstone identification is based on characterizing a sample’s color, transparency, birefringence, index of refraction, and specific gravity. Methods have now evolved to additionally rely on detecting key atomic structures and defects associated with these different materials [20–24]. Gemological laboratories use various spectroscopic techniques, often in tandem, including Fourier transform infrared (FTIR) and ultraviolet (UV)-visible absorption, Raman, photoluminescence (PL), and x-ray fluorescence spectroscopy [23]. Although laboratory services can be used for gemstone identification, they are not necessarily convenient, time-effective, or cost-effective. There is strong demand for inexpensive screening and identification instruments that can be used by non-specialized users such as gemstone dealers, manufacturers, and retailers, as well as small-scale gemological laboratories who lack advanced equipment or expertise.

The ideal screening device should be optimized to yield high true positive identification rates (accuracy) with zero or low false positive identification rates, yet have minimal restrictions on the types of applicable gemstones materials. To maximize flexibility, screening and identification instruments should be able to test both loose and mounted gemstones in a wide size and shape range and be operated at ambient lighting and temperature conditions. Automatic identification, rather than user data interpretation, is also desirable. Finally, the instrument’s testing speed, maintenance frequency, and purchase price define its operation costs. Although there are a range of gemstone screening devices available in the market, their advantages and limitations depend on the underlying technology and design. Oftentimes, the principles behind the instruments are not disclosed, complicating the comparison and evaluation of their applicability.

Considering colorless and near-colorless diamond screening (separating natural from laboratory grown), there are three main technologies available for non-specialized users in the trade. The first group of devices are based on differences in the absorption of natural and laboratory grown diamonds. Most common are those focusing on the UV-visible range. Natural diamond positive identification rates of ${\sim}51.6\%$ to 95.3% have been reported (meaning that the remaining natural diamonds were referred for further testing or incorrectly identified as synthetics or simulants) [25], partly depending on whether they are indirectly detecting the transmission of UV light or collecting and analyzing UV-visible absorption spectra. Some of these devices have been reported to pass laboratory grown diamonds as natural, and many cannot be used on sample sets that contain simulants. Other absorption-based screening devices collect and analyze infrared (IR) spectra and can achieve high (${\sim}95.9\%$) natural diamond positive identification rates while referring 100% of laboratory grown materials [25]. However, jewelry settings may restrict sample access for absorption measurements, potentially reducing their sensitivity or leading to tedious sample handling. Also, absorption spectroscopy-based technologies can be slow and expensive. A second common group of devices consider differences in fluorescence and phosphorescence between natural and laboratory grown diamonds by image analysis, in some cases relying on user interpretation. These systems have ${\sim}81.4\%$ to 99.7% natural diamond positive identification rates, but some suffer from natural diamond false positive identification rates of up to 19.6%, where synthetics are misclassified as natural [25]. Again, many of the devices in this group cannot be used with simulants. Also, the phosphorescence properties of some laboratory grown diamonds can be modified by treatment [26]. Finally, fluorescence- or PL-spectroscopy-based analysis is an alternative gemstone identification method [27,28]. Yet, simple systems that do automatic data interpretation are scarce. Typically, this analysis is done at gemological laboratories using costly and complex PL spectrometers equipped with multiple excitation laser wavelengths. Additionally, diamond samples are cooled to liquid nitrogen temperatures to maximize defect detection sensitivities. These requirements, coupled with undisclosed identification criteria, severely restrict their use.

This study presents a gemstone screening device based on fluorescence spectroscopy using a single UV light emitting diode (LED) for excitation. This device targets characteristic fluorescence features from gemstones, combined with automatic spectral analysis algorithms, to achieve accurate, rapid, and simple gemstone screening and identification, with flexibility to measure both loose samples and mounted jewelry pieces in standard office lighting and temperatures without additional sample preparation. The primary purpose of the device is to screen out laboratory grown diamonds and all types of simulants from natural colorless to near-colorless (${\rm D}$ to ${\rm J}$ grades) diamonds. Its potential application has been expanded with an additional algorithm to identify treated natural or laboratory grown pink diamonds. Its suitability to identify the mineral type of popular colored gemstones, such as corundum, spinel, beryl, and zoisite by automatic data interpretation has also been considered. The simple use, flexibility, low cost, and high accuracy of this instrument meets several jewelry industry screening device needs.

2. BACKGROUND ON GEMSTONE DEFECTS

Diamond is the only gem material that consists of only a single element—carbon atoms arranged in a tetrahedral arrangement. Due to its wide band gap (${\sim}5.5\;{\rm eV}$), it is transparent from the near-IR through the visible and UV range, meaning that a “prefect” diamond appears colorless. However, real diamonds actually contain a variety of crystalline atomic defects, which can be extended defects such as dislocations and slip planes (errors in the lattice), or atomic impurities where non-carbon elements (e.g., nitrogen) are presented. The presence of specific defects can be detected using spectroscopic methods such as absorption or PL spectroscopy, where electronic or vibrational transitions at these defects result in a measurable response by the spectrometer at specific wavelengths. Certain defect species may be active in either absorption or PL, or both, where PL may be the more sensitive technique. The absorption intensity of a defect is proportional to the concentration of that defect, yet the relationship between the PL intensity and the defect concentration is more complex, as it may involve both radiative and non-radiative energy relaxation mechanisms [29]. Hence, it is possible that the presence of certain defect species may decrease the emission efficiency of another, effectively quenching the PL. Defects, which are active in the visible range, and whose absorption spectra can produce color, are known as “color centers.” Diamond identification criteria is largely based on the analysis of defects using both absorption and PL techniques. Collins presents comprehensive reviews of defects in diamond and spectroscopic ways that they can be identified [6,16,30], while Eaton-Magaña and Breeding focus on the use of PL spectroscopy for gemological applications[28].

Nitrogen-related defects are commonly found in diamond and can exist in a wide range of configurations such as isolated nitrogen, aggregated (e.g., ${{\rm N}_{2}}$), or nitrogen-vacancy complexes (e.g., ${{\rm N}_{2}}{\rm V}$, ${{\rm N}_{3}}{\rm V}$, ${{\rm N}_{4}}{\rm V}$, etc., where ${\rm V}$ represents a vacancy) [6]. For the purpose of this study, the defect commonly known as the N3 defect, consisting of three nitrogen atoms surrounding a vacancy (${{\rm N}_{3}}{\rm V}$), is of interest [31,32]. The N3 defect has a zero phonon line (ZPL) at 415 nm, with a vibronic (sideband) structure at lower wavelengths in excitation and to higher wavelengths in emission, as illustrated in Fig. 1. N3 is present in the majority of natural diamonds, and, if sufficiently concentrated, the stones can show yellow color due to the defects’ absorption in the blue region of the visible spectrum. These diamonds are sometimes referred to as “cape” diamonds. Despite its frequency in natural diamonds, N3 is not formed in colorless and near-colorless chemical vapor deposition (CVD) and high pressure, high temperature (HPHT) laboratory grown diamonds in measurable concentrations, as the low nitrogen concentrations of the material (necessary to attain these colors), coupled with the synthesis growth temperatures and timescales, are not conducive to nitrogen aggregation [7,11,33,34]. The N3 defect structure, being restricted to diamonds, does not exist in diamond simulants. Therefore, the detection of the N3 defect can be used as a rapid screening method to distinguish natural diamonds from their laboratory grown and simulant counterparts, with PL spectroscopy (on average) providing enhanced sensitivity combined with testing flexibility [35]. However, as N3 can be present in both treated and untreated natural colorless to near-colorless color diamonds, it cannot be used for treatment screening.

 

Fig. 1. Excitation and emission spectra of the N3 (${{\rm N}_{3}}{\rm V}$) defect in diamond at room temperature. N3’s vibronic structure in excitation and emission are approximately symmetric about the zero phonon line at 415 nm. The 385 nm LED that is used in this study’s prototype device overlaps with the local maxima wavelength in the excitation spectrum.

Download Full Size | PDF

The main defects responsible for pink color in natural diamonds are currently unidentified defects associated with a broad absorption band at 550 nm [36–39] or nitrogen-vacancy defects, which have ZPLs at 575 nm (${{\rm NV}^{0}}$) and 637 nm (${{\rm NV}^{- }}$), with vibronic structures that absorb in the blue-to-green region of the visible spectrum [15,40]. The review article by Eaton-Magaña et al. presents examples of absorption and PL spectra for natural pink diamonds colored by both types of defects [15]. Only about 0.6% of natural pink diamonds are colored by ${{\rm NV}^{{0/} - }}$ defects. These extremely rare diamonds, known as “Golconda” diamonds in the gem trade, are characterized by pale color saturations. On the other hand, ${{\rm NV}^{{0/} - }}$ is responsible for the color in the vast majority of treated natural and laboratory grown pink diamonds [1,6,10,15–18]. Sufficiently high ${{\rm NV}^{{0/} - }}$ concentrations to introduce pink color can be artificially created in certain isolated nitrogen-containing diamonds by multi-treatment using (typically electron) irradiation followed by annealing at temperatures between $600 {-} {1200^ \circ}{\rm C}$. The isolated nitrogen could be present in natural or as-grown laboratory grown diamonds, or it could be produced in natural diamonds with aggregated nitrogen by HPHT treatment [5,41].

The popular non-diamond colored gemstones considered in this study are colored by trace element lattice impurities [19]. For example, the red color of rubies and green color of emeralds are attributed to the presence of chromium. Chromium is also an important cause of red fluorescence in many minerals, such as corundum, spinel, beryl, zoisite, topaz, and garnet, resulting in emissions near 700 nm [42]. Therefore, the luminescence spectra for these colored gemstones could potentially be used to effectively identify their corresponding mineral type.

3. EXPERIMENTAL DETAILS

The ideal light source for gemstone detection should efficiently excite the fluorescence signal from key defects in the tested sample. Conventionally, a mercury lamp’s 365 nm emission line has been used as the long-wave UV light source to study the fluorescence of gemstones [27]. When considering diamonds, the most common fluorescence color observed for natural samples is blue, associated with the presence of the N3 color center [43]. Visual inspection or color cameras are only able to detect the blue fluorescence in approximately 35% of natural diamonds when excited using this excitation, despite the prevalence of the N3 defect [43]. To improve the N3 fluorescence efficiency in diamonds, we have selected a 385 nm UV LED as the light source, coinciding with the local maxima in the N3 excitation spectrum, as presented in Fig. 1. Other excitation sources ranging from 365 to 405 nm would also be suitable; however, laser sources may be necessary to achieve comparable or improved N3 emission. In order to prevent accidental eye damage during operation, laser-based devices would require additional safety restrictions, increasing the cost and setting limitations on sample handling. The use of a 1R Class UV LED for this prototype provides a suitable inexpensive excitation source, which is safe to use under standard operating conditions, as long as additional magnifying optics are not used to look directly at the probe head.

A schematic layout of the fiber-based fluorescence spectrometer is presented in Fig. 2(a). A 385 nm LED generates the long-wave UV light used to excite the fluorescence from the gemstones. The fiber relays the UV light to the sample and collects and transmits the fluorescence signal to the detector. Finally, a spectrometer records, and a computer analyzes the fluorescence spectra from the sample, generating a screening or identification result based on a custom algorithm, discussed in Section 4.

 

Fig. 2. Optical design of the fluorescence spectrometer. (a) Schematic layout of the fluorescence spectrometer. (b) Shortpass filter set layout: a shortpass filter is placed after the UV LED to cut off the visible emission from the LED. (c) Longpass filter set layout: a longpass filter is placed before the spectrometer to cut off the UV emission from the LED. (d) Reflection fiber probe design: the light source fiber has six 200 µm core fibers, while a single 200 µm fiber is used for detection. These seven fibers merge at the probe tip, with the detector fiber located at the center.

Download Full Size | PDF

In order to improve the spectral quality, two filters are used to isolate the sample’s fluorescence spectrum from the light source emission spectrum, as presented in Figs. 2(b) and 2(c) [44]. Appropriately positioned longpass and shortpass filters cut off the primary excitation wavelength as well as longer wavelength components from the LED spectrum, respectively. As the optical filter design requires small incident angles to maintain the cutoff efficiency, lenses are used to collimate the light before the filters and focus the filtered light to the fiber.

A reflection probe with a 1.5 mm tip is used as the probe head. The probe has a 0.7 mm diameter fiber core bundle, suitable for measurement from gemstones with exposed surface areas greater than 0.9 mm. For a round brilliant faceted diamond, this requirement translates to a minimum size of 0.005 ct. As presented in Fig. 2(d), six fibers guide the UV light to excite the sample, and a single signal fiber relays the fluorescence signal to the detector.

The experimental prototype shown in Fig. 3 was built to test the concept of using room temperature fluorescence spectroscopy with a UV LED excitation for the screening of diamond and other gemstones. The light source for this setup was a 385 nm LED (M385L2, Thorlabs) driven by a driver (LEDD1B, Thorlabs). Two 10.5 mm focal length aspheric condenser lenses (ACL1210U-A, Thorlabs) were installed before and after the 390 nm shortpass filter (390SP-OD4, Omega Optical) to collimate, filter, and then couple the selected UV source into the reflection fiber probe (FCR-7IR200-0.40-${1.5} \times {10}$, Avantes). The same aspheric condenser lens was used to collimate the fluorescence signal into a 409 nm longpass filter (FF02-409/LP-25, Semrock). A 30 mm focal length plano-convex lens was use to couple the filtered fluorescence signal into a 200 µm multimode signal fiber (M92L01, Thorlabs). All of the optics were mounted and aligned using a 30 mm cage system (Thorlabs). The fibers were connected through subminiature version A (SMA) connectors (SM05SMA, Thorlabs). A spectrometer was used for signal detection instead of a photodiode or a color camera, as spectral analysis can be used to distinguish a gemstone’s fluorescence from ambient light and other optical background signals. The spectrometer (AvaSpec-Mini 2048L, Avantes) had a detection wavelength range from 400 to 900 nm with 1.2 nm spectral resolution. The entire setup was mounted on a ${6}\;{\rm in}. \times {12}\;{\rm in}.$ aluminum optical breadboard (MB612F, Thorlabs) for portable applications.

 

Fig. 3. Experimental prototype of the fluorescence spectrometer.

Download Full Size | PDF

An optical power sensor (S120VC, Thorlabs) was used to calibrate the radiant flux of the UV source at the probe head. The output power of the LED was set to 5 mW after warming up the LED for 30 min. To test the cutoff efficiency of the filter setup, the LED signal transmission to the spectrometer was measured by illuminating an aluminum plate. Light transmission through the filters could be detected, yet was sufficiently weak compared to the fluorescence signal of most samples to be easily discounted by the analysis algorithm. The samples’ fluorescence intensities can vary significantly, and the detection efficiency can be affected by both the probe incidence angle and the distance between the probe and sample. The fiber properties are also important, with the fiber’s core diameter, numerical aperture, and attenuation impacting the signal intensities. Strong fluorescence signals may saturate the detector, interfering with spectral analysis, whereas weak fluorescence signals may lead to poor signal-to-noise, hindering peak detection. To resolve this issue, an auto-exposure function was applied to dynamically adjust the spectrometer’s integration time to intervals between 2 ms to 3 s.

To test the performance of the prototype and the diamond screening software, we measured 9560 ${\rm D}$ to ${\rm J}$ color samples including 7685 natural diamonds, 1864 laboratory grown diamonds, and 11 diamond simulants, with 2765 pink-hued diamonds, including 2045 untreated natural diamonds, 706 multi-treated diamonds, and 14 laboratory grown diamonds. Simulants tested were synthetic spinel, topaz, cubic zirconia (CZ), zircon, synthetic rutile, strontium titanate, gadolinium gallium garnet (GGG), synthetic sapphire, synthetic moissanite, yttrium aluminum garnet (YAG), and a synthetic spinel/strontium titanate doublet. There were 775 non-diamond colored gemstones tested: 417 corundums (including both rubies and sapphires), 99 beryls, 98 spinels, 39 zoisites, 69 topazes, and 53 garnets. The sample suite included gemstones from GIA’s research sets, the Dr. Edward J. Gübelin Gem Collection, as well as client-submitted stones from our grading laboratory operations, all previously conclusively identified by other methods. A jewelry piece that included 1 HPHT laboratory grown and 19 natural melee-sized diamonds was used to assess the prototype’s suitability for mounted gemstone testing. Thousands of additional samples have been tested, but are excluded from this study as statistics were not recorded. To maximize the detected signal intensity, the probe head was placed in contact with a polished surface for each of the loose or mounted samples. For the latter group of stones, the table facet was the most easily accessible. Fluorescence integration times ranged from 2 ms to 1 s for diamond and 2 ms to 3 s for colored gemstones. Corresponding dark and room light spectra were subtracted from the measurements. All measurements were conducted at room temperature. A spectral analysis algorithm was developed to automatically screen the samples based on the existence or absence of characteristic fluorescence peaks.

4. RESULTS AND DISCUSSION

A. Natural Colorless and Near-Colorless Diamond Screening

Examples of characteristic fluorescence spectra for natural, HPHT-, and CVD-grown diamonds are presented in Fig. 4. Only the spectra for natural diamonds showed N3 fluorescence, with the ZPL at 415 nm and a vibronic sideband (extending from the ZPL to ${\sim}565\;{\rm nm}$ [31]). Conversely, the spectra for laboratory grown diamonds either were inert or showed characteristic features other than N3 fluorescence. As they are chemically distinct from diamond materials, simulants cannot contain the N3 defect nor show the associated fluorescence signature. Figure 5 presents the fluorescence spectra for diamond simulants including synthetic spinel, zircon, GGG, synthetic sapphire, and YAG. Conveniently, these fluorescent simulants also did not emit at 415 nm from other unrelated sources. Alternative diamond simulants tested such as CZ, colorless topaz, synthetic rutile, strontium titanate, and synthetic moissanite did not show detectable fluorescence using our prototype (spectra not shown).

 

Fig. 4. Typical fluorescence spectra for natural, HPHT-, and CVD-grown diamonds. N3 fluorescence at 415 nm was only detected for natural diamonds. The weak signal at 410 nm observed for the latter spectrum originates from leakage of the UV LED signal through the filter and can be discounted.

Download Full Size | PDF

 

Fig. 5. Typical fluorescence spectra for fluorescent diamond simulants, including synthetic spinel, zircon, gadolinium gallium garnet (GGG), synthetic sapphire, and yttrium aluminum garnet (YAG).

Download Full Size | PDF

Additionally, a small proportion of natural diamonds containing high concentrations of aggregated nitrogen defects including N3 (as observed by absorption spectroscopy) do not necessarily have detectable N3 fluorescence. This stems from the presence of nitrogen pair defects, known as A-centers, which can quench N3 fluorescence [45]. In our experiments, we found that more than 50% of these N3-inert diamonds showed fluorescence that visually appeared “white” or “yellow” during illumination (and are referred to as such in the trade). The associated asymmetric broad emission bands, presented in Fig. 6, peak at ${\sim}525\;{\rm nm}$ (green) and ${\sim}560\;{\rm nm}$ (greenish yellow). These characteristic spectra could also be used to conclude that these samples are natural. Unlike the distinctive N3 signal, the white or yellow florescence bands may overlap with the emission of white LED room light. An accurate room light spectral subtraction helps distinguish the white and yellow bands from background lighting conditions.

 

Fig. 6. “White” and “yellow” fluorescence from natural diamonds, with broad asymmetric emission bands peaking at ${\sim}525$ and ${\sim}560\;{\rm nm}$, respectively. The fluorescence color was defined by visual evaluation.

Download Full Size | PDF

Based on the characteristic spectral features outlined above, an automatic diamond screening algorithm for colorless and near-colorless diamonds was developed. It used the presence of N3, white, or yellow fluorescence bands to identify natural diamonds and separate them from laboratory grown diamonds and simulants. The algorithm defined N3 fluorescence by the detection of a 415 nm peak and the vibronic peaks at 428 and 439 nm. The white and yellow fluorescence bands were confirmed by measuring the signal’s maximum peak position, intensity, and width and comparing them to defined parameters for known samples. Diamonds showing N3, white band, or yellow band fluorescence were identified by the algorithm as natural diamonds, whilst those who did not were referred for further testing. Table 1 summarizes the results for colorless and near-colorless diamond screening. The algorithm detected natural diamond fluorescence in 99.7% of natural diamonds equal to or larger than 0.30 ct and in 94.5% of natural diamonds smaller than 0.23 ct (melee-sized). On the other hand, none of 1231 HPHT- and 633 CVD-grown diamonds nor simulants showed these features. Hence, for the samples tested, the overall natural diamond positive identification rate was 97.0%, while 100.0% of laboratory grown diamonds and simulants were referred (0% natural diamond false positive identification rate). The absence of N3 fluorescence in the spectra collected for colorless and near-colorless laboratory grown diamonds under these experimental conditions are consistent with our understanding of defect formation in both HPHT- and CVD-grown diamonds as well as with observations reported using higher sensitivity techniques, such as low temperature PL spectroscopy [7,11,33,34].

Table 1. Data Collection for Colorless and Near-Colorless (${\rm D}$ to ${\rm J}$ Grades) Diamond Screening

View Table | View all tables in this article

Diamonds are often classified and segregated into “diamond type” groups (types IaA, IaB, IaAB, IIa, and IIb) based on the presence and absence of nitrogen- and boron-related FTIR absorption features. The different diamond types are defined in the Supplement 1. Table 2 summaries the natural diamond positive identification rate for different diamond types for a suite of 277 natural diamonds with color grades ranging from ${\rm D}$ to ${\rm Z}$. Based on this data, 100.0% of type Ia saturated, 100.0% of type IaB, 96.2% of type IaAB, and 84.6% of type IaA diamonds had detectable fluorescence and could be identified as natural diamonds. On the other hand, fluorescence was not detected for 51.2% of type IIa diamonds nor any type IIb diamonds. These nitrogen-poor types are rare and combined account for less than 2% of natural gem diamonds [46–49].

Table 2. Different Types of Natural ${\rm D}$ to ${\rm Z}$ Color Diamonds and the Corresponding Natural Diamond Positive Identification Rates for Our Prototype^a

View Table | View all tables in this article

 

Fig. 7. Typical fluorescence spectra for pink natural color, multi-treated color, CVD-, and HPHT-grown diamonds. The natural pink diamond spectrum shows N3 fluorescence at 415 nm without any detectable ${{\rm NV}^{0}}$ fluorescence at 575 nm, while multi-treated color, CVD-, and HPHT-grown pink diamond spectra have detectable ${{\rm NV}^{0}}$ fluorescence.

Download Full Size | PDF

B. Natural Pink Diamond Screening

As explained in Section 2, the defects responsible for pink color in natural and multi-treated diamonds differ [15]. Figure 7 shows characteristic fluorescence spectra for natural color, multi-treated color, HPHT-grown, and CVD-grown pink diamonds. As shown in the figure, the latter three groups had a detectable ${{\rm NV}^{0}}$ fluorescence peak at 575 nm along with vibronic structure, while most natural pink diamonds primarily emitted N3 fluorescence without sharp spectral features between 550 and 650 nm.

Strong orange–red fluorescence associated with ${{\rm NV}^{0}}$ centers was ubiquitous in the spectra for the multi-treated pink natural and laboratory grown diamonds. Green fluorescence contributions from H3 (${{\rm NVN}^{0}}$) and H4 (${{\rm N}_{4}}{{\rm V}_{2}}^{0}$) defects, or an absorption peak (seen as a dip in fluorescence) at 637 nm attributed to the negatively changed NV center (${{\rm NV}^{- }}$), were also observed in some of these samples. Under our prototype at room temperature, ${{\rm NV}^{0}}$ fluorescence has a strong fluorescence side band, which overlaps with the ${{\rm NV}^{- }}$ emission peak at 637 nm. The negative peak/dip in the spectrum (Fig. 7) indicates that the emitted light at that wavelength was strongly absorbed by the sample’s ${{\rm NV}^{- }}$ centers. As a result, for this setup, the detection of the emission peak from ${{\rm NV}^{0}}$ is a better candidate than that from ${{\rm NV}^{- }}$ for pink diamond screening. Similarly, although the H3 and H4 defects have potential uses for treatment detection, they were not included in our analysis algorithm, as their features could be distorted by overlapping fluorescence signals from other sources. N3 fluorescence was observed for both treated and untreated natural diamonds.

The pink diamond screening algorithm considered N3 and ${{\rm NV}^{0}}$ fluorescence to confirm the origin of the pink color in diamond. The presence of N3 was defined in the same way as for the colorless and near-colorless screening algorithm; ${{\rm NV}^{0}}$ was detected using the 575 nm peak. To identify a sample as a natural untreated pink diamond, the sample’s fluorescence spectrum had to meet two conditions: showing detectable N3 and simultaneously no ${{\rm NV}^{0}}$ emissions. All other samples were referred for additional testing by the algorithm, without further distinguishing between those that were treated naturals or laboratory grown.

Table 3 summarizes the data collection for the pink diamond screening tests. N3 fluorescence was detected for 99.0% of the natural untreated pink diamonds larger than 0.30 ct and 100.0% of those samples smaller than 0.23 ct (melee-sized). Only 0.2% of natural color untreated pink diamonds showed detectable ${{\rm NV}^{0}}$ fluorescence and were referred. Meanwhile, ${{\rm NV}^{0}}$ fluorescence was ubiquitous for the multi-treated natural and synthetic diamonds. The overall natural untreated pink diamond positive identification rate was 99.8%, while 100.0% of multi-treated natural or synthetic pink diamonds were referred (0% natural untreated pink diamond false positive identification rate).

Table 3. Data Collection Collected for Using the Pink Diamond Screening Algorithm

View Table | View all tables in this article

It is important to note that there is an alternative treatment procedure for creating pink color in natural diamonds. HPHT annealing may be used to remove unattractive brown color, potentially revealing an underlying pink color in certain type IIa or IaB diamonds, unrelated to NV centers [37,50,51]. However, this group of treated pink diamonds are extremely rare due to the scarcity of the starting material [52]. The ${{\rm NV}^{0}}$ concentration for this type of sample would likely be too low to detect by this prototype. Yet, some of these samples may show detectable N3 centers. Thus, they would not be reliably referred by our setup.

C. Colored Gemstone Identification

Typical fluorescence spectra of corundum (rubies and sapphires), spinel, beryl, zoisite, topaz, and garnet collected with our device are shown in Fig. 8. The most common fluorescence features detected were associated with trivalent chromium ions (${{\rm Cr}^{{3} +}}$) substituting for cations in the host mineral. Due to the differing host chemical structures, the chromium ions produce sharp emission peaks at characteristic positions unique to each mineral [53]. We evaluated whether these distinctive spectral features could be used to separate the different colored gemstones. Analysis of spectra collected with this instrument showed the presence of these identifying features in 94.5% of corundum, 97.0% of beryl, 93.9% of spinel, and 84.6% of zoisite samples tested. Yet, only 30.4% and 20.8% of topaz and garnet samples showed the characteristics features seen in Figs. 8(e) and 8(f), respectively. The complexity of colored gemstones, including their range of colors (and chromophores), mean that it is more appropriate to state identification rates for individual colors for each mineral rather than a general identification rate for each mineral, which would be biased by the sample selection. The detailed data analyses for each type of mineral, including features that were incorporated into an automatic colored gemstone identification algorithm, are presented in Supplement 1.

 

Fig. 8. Typical fluorescence spectra of common colored gemstones in the trade market: (a) corundum, including ruby and sapphire, (b) spinel, (c) beryl, (d) zoisite, (e) topaz, and (f) garnet.

Download Full Size | PDF

A suitable application for this device would be to help identify or rule out similarly colored gemstones. For instance, the most likely mineral options for red gemstones (like the originally misidentified Black Prince’s Ruby and the Timur Ruby) would be corundum (ruby), spinel, garnet, and topaz. Our results indicate that 100% of red corundum, spinel, and topaz samples would be positively identified, yet garnets would not. Among blue gemstones, 89.8%, 85.7%, and 90.0% of corundum (sapphire), spinel, and zoisite samples would be identified, respectively, while topaz would not. For green gemstones, this device could identify the majority of beryl (emerald, 96.9%) and zoisite (100%) samples. Low detection rates for garnets suggest that this device would be of limited use for their analysis. The gemstones used in this study were representative of what is observed in the gem trade, and no attempts were made to further separate the materials based on treatment, synthesis, or country of origin. Importantly, no samples were incorrectly identified as different minerals (0% false positive identification rates).

D. Mounted Jewelry Measurement

A pendant, shown in Fig. 9, was tested to verify the suitability of the screening device prototype for mounted jewelry testing. The pendant included 19 natural diamonds and one HPHT-grown diamond (colorless to near-colorless). Figure 9(b) illustrates the fiber probe being used to detect the fluorescence signal of a single stone in the mounted piece. The probe successfully confined the UV excitation to the test stone, as well as limited the detection of fluorescence to that originating from that stone. Of the natural diamonds, 100% were identified using our prototype, with similar signal intensities as those observed for unmounted samples. In the gem trade, fluorescence is typically evaluated by visual inspection using a mercury lamp’s long-wave UV 365 nm emission line [27]. Figure 9(c) shows a fluorescence image of the pendant under such a light, demonstrating that our prototype was sufficiently sensitive to detect fluorescence signals even for the natural diamonds that appeared inert to long-wave UV. The HPHT-grown diamond was referred by the device. Although this example only considers diamond materials, we can infer that it would also apply to mounted colored gemstones, as their identification is also based on fluorescence spectroscopy.

 

Fig. 9. Mounted jewelry piece measurement. (a) A pendant including 20 melee diamonds was used to test the prototype’s application for mounted jewelry. The fourth diamond from the left was a HPHT-grown diamond, as indicated. (b) The fiber probe could easily test the individual mounted melee diamonds. (c) Fluorescence image of the pendant under long-wave UV illumination by a mercury lamp’s 365 nm emission. The natural diamonds showed a range of fluorescence intensities, while the HPHT-grown diamond appeared inert. (d) Once the UV source was turned off, the HPHT-grown emitted greenish-blue phosphorescence, a common characteristic for HPHT-grown diamonds[2].

Download Full Size | PDF

5. CONCLUSIONS

We have developed a prototype gemstone screening device suitable for non-expert users based on UV fluorescence spectroscopy, experimentally demonstrating its functionality and accuracy. To optimize the fluorescence excitation and detection efficiencies of relevant identification features, the device uses a 385 nm LED as the light source, two filters to isolate the excitation and fluorescence signals, and a spectrometer to capture the fluorescence signal from the gemstones. A fiber probe segregates and effectively transmits both the excitation and sample-emitted signals between the device and test samples. This design allows testing of both loose and mounted gemstones (faceted or rough) that have a diameter greater than 0.9 mm, expanding the instrument’s suitability to encompass a wide range of sectors within the gem and jewelry trade. The system, including proprietary analysis algorithms, was demonstrated to confidently identify colorless to near-colorless natural diamonds and refer laboratory-grown diamonds and simulants, as well as separate natural from multi-treated treated natural and synthetic pink diamonds. A 0% natural diamond false positive rate was achieved for both groups of samples. Its application for colored gemstone identification was also tested for popular minerals such as corundum (ruby and sapphire), beryl (including emeralds), spinel, zoisite (also known as tanzanite), topaz, and garnets. Analysis of the samples’ fluorescence spectra allowed the identification of most corundum, beryl, spinel, and zoisite samples. The lower fluorescence intensities from the remaining gemstone minerals led to lower positive identification rates. Notably, none of the colored gemstones were incorrectly identified as other minerals.

Future research will focus on the feasibility and development of a screening protocol for other treated fancy color diamonds and rare colored gemstones. We note that the definition and techniques of synthetically produced diamonds and colored gemstones, as well as treatments to modify the quality of both natural and man-made materials, are continuously evolving. Hence, further studies will also target the latest technologies and corresponding effects for gemstone identification.