Ali Haiderab,
Mehmet Yilmazb,
Petro Deminskyib,
Hamit Erenab and
Necmi Biyikli*c
aInstitute of Materials Science and Nanotechnology, Bilkent University, Ankara 06800, Turkey
bUNAM – National Nanotechnology Research Center, Bilkent University, Ankara 06800, Turkey
cElectrical and Computer Engineering Department, Utah State University, Logan, UT 84322, USA. E-mail: n.biyikli@aggiemail.usu.edu
First published on 1st November 2016
Here, we report nano-patterning of TiO2 via area selective atomic layer deposition (AS-ALD) using an e-beam patterned growth inhibition polymer. Poly(methylmethacrylate) (PMMA), polyvinylpyrrolidone (PVP), and octafluorocyclobutane (C4F8) were the polymeric materials studied where PMMA and PVP were deposited using spin coating and C4F8 was grown using inductively coupled plasma (ICP) polymerization. TiO2 was grown at 150 °C using tetrakis(dimethylamido) titanium (TDMAT) and H2O as titanium and oxygen precursors, respectively. Contact angle, scanning electron microscopy (SEM), spectroscopic ellipsometry, and X-ray photoelectron spectroscopy (XPS) measurements were performed to investigate the blocking/inhibition effectiveness of polymer layers for AS-ALD of TiO2. TiO2 was grown with different numbers of growth cycles (maximum = 1200 cycles) on PMMA, PVP, and C4F8 coated substrates, where PMMA revealed complete growth inhibition up to the maximum number of growth cycles. On the other hand, PVP was able to block TiO2 growth up to 300 growth cycles only, whereas C4F8 showed no TiO2-growth blocking capability. Finally, mm-, μm-, and nm-scale patterned selective deposition of TiO2 was demonstrated exploiting a PMMA masking layer that has been patterned using e-beam lithography. SEM, energy-dispersive X-ray spectroscopy (EDX) line scan, EDX elemental mapping, and XPS line scan measurements cumulatively confirmed the self-aligned deposition of TiO2 features. The results presented for the first time demonstrate the feasibility of achieving self-aligned TiO2 deposition via TDMAT/H2O precursor combination and e-beam patterned PMMA blocking layers with a complete inhibition for >50 nm-thick films.
Controlling the lateral dimensions of thin films by patterning is pivotal in microelectronics industry due to ever-increasing trend towards further miniaturization of device feature sizes.4,5 Conventionally, thin film patterning is achieved by photolithography which includes several processing steps such as resist spinning, UV exposure, resist development, and film etching. ALD processes, in which film nucleation critically relies on the surface chemistry between gaseous precursors and the solid surface, provide an attractive opportunity for performing area-selective deposition by chemically modifying the substrate surface. Local modification of substrate surface opens up possibilities to achieve lateral control over film growth in addition to robust thickness control during ALD process.6–11 Area-selective ALD (AS-ALD) might pave the way for low-temperature self-aligned nanoscale device fabrication by reducing or eliminating lithography/etch process steps and minimizing hazardous reagent use. Taking these significant advantages into consideration, the efforts of developing reliable and effective AS-ALD recipes have attracted considerable interest in recent years. ALD-enabled nano-patterning has been classified under two broad categories, one with area-activated agents and the other with area-deactivated blocking/inhibition layers.7,9–34 So far, majority of the AS-ALD studies have been performed using area-deactivated approach where mostly self-assembled monolayers (SAMs) are utilized as the growth-blocking layers by covering the chemically reactive sites on the substrate and exposing non-reactive groups.7,9,23,27,29–40 Alkyl silanes e.g., alkyl trichlorosilanes, alkyl triethoxysilanes, etc. have been exploited as mono-layered surface modifiers to block ALD nucleation of various metal oxide thin films and metallic nanoparticles/thin films.10,13,36,37,41–48 In this strategy, chlorosilane compounds chemically react with hydroxyl sites on the substrate surface and expose only unreactive alkyl groups on the surface which serve as effective ALD nucleation preventing agents. Although promising, this approach depends critically on the availability of defect-free SAM blocking layers, otherwise the defects in SAM act as nucleation centers leading to reduced selectivity and eventually non-selective growth. Moreover, preparing defect-free SAMs is not easy and generally takes extremely long synthesis times (up to 48 h).23,31,41,48 Even with a decent quality SAM coating, growth selectivity might still be limited to a few nanometers of film thickness. In addition, patterning of SAMs has generally been attained using non-standard lithographic techniques such as micro-contact printing which further makes it a laborious task to obtain defect-free SAMs. Such a slow and rather unreliable masking process may undermine the capability of AS-ALD process as a straight-forward, fast, and reliable technique for potential use in high-volume manufacturing.
Overcoming the limitations associated with SAM-based mask layers require the production of easily patterned, non-reactive, and defect-free blocking layer materials. Polymer films present an alternative way to prepare defect free masking layers which physically screen the active sites on the substrate and enable AS-ALD process.24,49–51 Indeed, polymer films with several critical advantages including quick and easy coating, defect free film quality, and ease of patterning have been implemented in majority of the lithographic patterning processes. In this scenario, if one can identify a polymer or a group of polymers that are unreactive towards ALD precursors which can also be easily patterned and removed after the growth, then that polymer film can be potentially used as a blocking layer to achieve AS-ALD process. Such a self-aligned AS-ALD approach to obtain a directly patterned structure of a desired ALD film may avoid additional etching and lift-off processes associated with regular lithography-based patterning methods.
AS-ALD of TiO2, CeO2, ZnO, N-doped ZnO, Ru, Rh, and Pt have been demonstrated using various polymer layers as growth inhibitor.24,25,31,49–54 ALD-grown films might start nucleating on the polymer blocking layer after a certain number of ALD-cycles; patterning of such films are demonstrated via conventional lift-off processes. Al2O3, TiO2, ZnO, ZrO2, HfO2, CeO2, and Co have been patterned using polymer layers as lift-off resist films.55–59 In most of these studies poly(methyl methacrylate) (PMMA) or polyvinylpyrrolidone (PVP) have been utilized as either blocking or lift-off layers. Both polymers feature ease in coating, compatibility with conventional patterning techniques, and rather simple removal after the growth. Recently PMMA has also been utilized as a chemical sponge in sequential infiltration synthesis (SIS) technique to achieve AS-ALD of Al2O3.60
Blocking capability for area-selective deposition might depend not only on the type of blocking polymer materials used, but also on the specific ALD process conditions (growth recipes) such as employed precursors and doses, unit cycle and cumulative process time, reactor pressure, substrate temperature, etc.11,61 AS-ALD of TiO2 layers have been carried out previously using PMMA as blocking layer with titanium tetrachloride (TiCl4), titaniumisopropoxide Ti(OiPr)4, and titaniumethoxide Ti(OMe)4 as titanium precursor sources.24,25,53,59 Among these studies, successful AS-ALD results were achieved using Ti(OiPr)4 and Ti(OMe)4 precursors, both exhibiting effective growth inhibition on PMMA surfaces. On the other hand, TiO2 growth was observed on PMMA for TiCl4 precursor and therefore, patterning was performed using routine lift-off method. Thin film patterning of TiO2 in these studies was accomplished on a μm PMMA pattern defined using either optical or thermal probe based lithography methods. However, with the continuous downward scaling of electronic devices, self-aligned area selective ALD using a nano patterning scheme such as e-beam lithography is highly imperative. Adoption of selective deposition approaches in device fabrication also requires those thin film growth precursors which are completely unreactive towards growth inhibition layers in order to provide thickness independent selectivity. Keeping all these factors in mind, a continuous exploration for most appropriate growth precursor and inhibition layer that can be patterned at nanoscale is required. Towards this goal, for the first time, we report nano-patterning of TiO2 using tetrakis(dimethylamido)titanium (TDMAT) via AS-ALD using an e-beam patterned growth inhibition polymer which has been selected among a set of polymers. At first, we present a detailed investigation to determine the efficacy of PMMA, PVP, and octafluorocyclobutane (C4F8) polymeric blocking layers for AS-ALD of TiO2 harnessing TDMAT and H2O as titanium and oxygen precursors, respectively. PMMA and PVP were deposited using spin coating and C4F8 was deposited using inductively coupled plasma (ICP) polymerization. Contact angle, scanning electron microscope (SEM), spectroscopic ellipsometer, and X-ray photoelectron spectroscopy (XPS) measurements were performed to determine the most compatible polymer layer for AS-ALD process of TiO2. Finally, μm and nm-scale self-aligned growth of TiO2 has been performed using e-beam lithography of PMMA layer. SEM, energy-dispersive X-ray spectroscopy (EDX) line scan, EDX elemental mapping, XPS line scan, and transmission electron microscope (TEM) were employed to characterize the self-aligned deposition and patterning efficiency of TiO2.
Surface morphologies of the PMMA, PVP, and C4F8 films grown on Si (100) were examined by AFM and shown in Fig. S1(a)–(c).† All samples revealed smooth morphologies with the following root-mean-square (Rms) surface roughness values; PMMA/Si = 0.534 nm, PVP/Si = 0.158 nm, and C4F8/Si = 0.212 nm. PMMA film also revealed ∼5–6 nm deep pinholes at few places on the sample (inset Fig. S1(a)†). Fig. 1 shows the variation in contact angle and thickness of TiO2 with the increase in number of growth cycles on C4F8, PMMA, PVP, and Si(100). As PVP is soluble in water and other polar solvents, contact angle measurements using water as a solvent would not provide accurate results. Hence, contact angle measurements were only performed on C4F8, PMMA, and Si(100). Initial contact angle of C4F8, PMMA, and OH rich Si(100) was measured as 114°, 74°, and 0°, respectively. XPS analysis (Fig. S2†) showed that C4F8 is a mixture of fluorocarbons such as C-CF, CF, CF2, and CF3. The film is believed to be formed by the fragmentation of C4F8 monomers by plasma and dissociation of CFx radicals.62 Fluorocarbons are known to impart relatively high hydrophobicity to the desired surface. ICP-polymerized C4F8 coatings showed a contact angle of 114° which confirmed this hydrophobic nature. Contact angle of Si(100) and C4F8 samples reached to ∼35° as soon as they were exposed to 100 cycles of TiO2 growth. With further increase in TiO2 growth cycles, contact angle rises again and stabilizes around ∼62–63° till 1200 cycles. On the other hand, PMMA exhibits quite stable contact angle values around ∼73°, almost independent of the number of TiO2 ALD cycles. The fact that contact angle of PMMA doesn't change with TiO2 growth cycles suggests that PMMA is efficiently blocking TiO2 film growth. To confirm this observation, ellipsometric film thickness measurements were carried out. Fig. 1b shows the evolution of TiO2 thickness on different surfaces as a function of ALD-growth cycles. As anticipated with conventional ALD growth processes, a linear increase in thickness of TiO2 is observed on Si(100) with a GPC of ∼0.5 Å. TiO2 thickness increase on C4F8 is also linear and nearly matches with the TiO2 growth rate on Si(100), which indicates that the initially hydrophobic plasma polymerized C4F8 layer is rather ineffective in blocking TiO2 growth. On the other hand, no growth of TiO2 is observed on PVP layers up to 300 cycles, while a very thin TiO2 layer (∼1.29 nm) is detected at 400 cycles, signaling the nucleation initiation at this growth stage on PVP coatings. With the further increase in ALD cycles beyond 400, TiO2 eventually nucleates on PVP surface, where after the growth rate becomes similar as on Si(100).
![]() | ||
Fig. 1 Variation in (a) contact angle and (b) thickness of TiO2 with number of growth cycles on PMMA, PVP, C4F8 coatings, and reference Si(100) substrate. |
This result suggests that PVP surface is successful in blocking/delaying the TiO2 growth for more than 300 cycles which corresponds to an effective film thickness of ∼15 nm on Si surface. For PMMA-coated samples, we have observed that TiO2 doesn't nucleate on PMMA surface at all, and no film growth is detected up to 1200 ALD cycles. These results indicate that PMMA is the most effective surface for TiO2 growth inhibition among the coatings/surfaces studied.
Previous studies on AS-ALD established a direct correlation between surface energy and water contact angle to the growth inhibition ability of SAMs. In a case study of AS-ALD of HfO2 with SAMs, it has been reported that only ODTS with a sufficiently high water contact angle is effective in blocking nucleation. Short or branched chained SAMs with low water contact angle were not able to inhibit nucleation of HfO2.63 In another study of AS-ALD of TiO2 with mixed SAM surfaces, it was observed that extent of nucleation increases with decreasing surface energy or water contact angle of SAM surfaces.64 Higher contact angle of SAM surfaces was only possible for well-packed SAM structures and degree of packing is an important parameter in AS-ALD processes using SAMs. High degree of packing prevents the ALD precursor access to reactive sites on Si substrates while superior hydrophobicity of SAMs prohibits the chemisorption of water which in turn blocks the nucleation of desired material. On the basis of these previous studies, one would expect C4F8 to show the highest nucleation delay due to its hydrophobic character and initially high contact angle. However, contact angle and spectroscopic ellipsometer measurements contradicts this prediction and show that TiO2 nucleates on C4F8 with relative ease, showing almost no nucleation delay. PMMA, on the other hand, with a water contact angle significantly smaller then C4F8, is quite effectively blocking TiO2 growth. Therefore, these results indicate that attaining successful AS-ALD depends on mainly two critical factors: (i) polymer blocking layer should be able to provide a sufficient barrier for ALD precursors to reach active sites on the surface, (ii) undesired reactions between inhibition layer and the ALD precursors must be avoided. In order to perform elemental quantification, XPS measurements were conducted on TiO2 grown on PMMA, PVP, Si(100), and C4F8 as a function of ALD cycles up to 1200. Fig. 2 shows XPS survey scans from TiO2 grown on PMMA and PVP coatings. Only C1s and O1s peaks are detected from PMMA surface till 1200 cycles of TiO2 growth. Absence of any Ti peak confirms that PMMA successfully abstain itself from TiO2 nucleation. Only C1s and O1s peaks are detected on PVP up to 300 cycles of TiO2 growth, where after Ti2p peak is observed. Fig. 3 shows XPS survey scans from TiO2 grown on C4F8 and Si(100). C1s, Ti2p, and O1s peaks are observed from TiO2 grown on C4F8/Si, while F1s peak is observed from the same substrate only with 100 cycles of TiO2 growth. As anticipated, TiO2 growth on Si(100) reveals the peaks of C1s, Ti2p, and O1s regardless of number of ALD cycles. These results confirm the rather quick nucleation of TiO2 and ineffective blocking behavior of both Si and C4F8-coated surfaces.
Quantification of Ti in terms of atomic percentages (at%) from survey scans from all four surfaces studied is summarized in Table 1.
Number of ALD cycles | Ti at% on C4F8/Si | Ti at% on Si(100) | Ti at% on PMMA/Si(100) | Ti at% on PVP/Si(100) |
---|---|---|---|---|
100 | 14.06 | 21.23 | 0 | 0 |
200 | 21.69 | 23.95 | 0 | 0.92 |
300 | 22.73 | 23.32 | 0 | 1.15 |
400 | 23.23 | 23.33 | 0 | 17.25 |
600 | 22.15 | 25.32 | 0 | 24.93 |
800 | 24.82 | 25.23 | 0 | 24.82 |
1000 | 23.58 | 25.21 | 0 | 23.52 |
1200 | 24.52 | 24.21 | 0 | 24.25 |
These XPS survey scan results provide an excellent correlation with contact angle and ellipsometer measurements and approve the following important conclusions: (i) PMMA successfully blocks/inhibits the TiO2 deposition for at least 1200 growth cycles, which is equivalent to a blocking film thickness of ∼55 nm (ii) PVP blocks TiO2 growth up to 300 ALD cycles and further increase in growth cycles eventually leads to nucleation of TiO2 on PVP, (iii) C4F8 is unable to inhibit TiO2 nucleation and growth, despite its higher initial contact angle.
Another important observation was the decrease in PMMA film thickness with number of TiO2 ALD cycles, which is presented in Table 2. We had chosen the substrate temperature as 150 °C which is slightly below the glass transition temperature (Tg = 108–167 °C) of PMMA.65 Decrease in PMMA thickness might be partly due to residual solvent removal during excessively long growth periods. In addition to inherent unreactive nature of PMMA, this slight decrease in thickness of PMMA can possibly aid in achieving a better selectivity.
Number of TiO2 cycles | Thickness of PMMA |
---|---|
0 | 43 nm |
100 | 42.91 |
200 | 41.516 |
300 | 37.561 |
400 | 35.45 |
600 | 33.99 |
800 | 32.84 |
1000 | 27.40 |
1200 | 23.96 |
Fig. 4 shows the Ti2p high resolution (HR)-XPS scans obtained from TiO2 grown on PMMA and PVP with various number of ALD cycles. In accordance with the observations made by XPS survey scans, no Ti2p peak is detected from PMMA samples regardless of the number of ALD-growth cycles and on PVP up to 300 ALD cycles. Ti2p3/2 and Ti2p1/2 peaks are observed at a binding energy of 458.99 and 464.80 eV for 400 and 600-cycle TiO2 respectively, grown on PVP/Si. These peaks are in agreement with the literature reports where Ti2p3/2 and Ti2p1/2 peaks are typically observed from TiO2 at a binding energy value of 458.5–458.9 and 463.7–464.2 eV, respectively, which are assigned to the distinct Ti4+ chemical state of Ti in TiO2.66,67 Same Ti2p peaks are observed for PVP samples with TiO2 ALD cycle numbers higher than 600.
![]() | ||
Fig. 4 HR-XPS survey scans of Ti2p obtained from TiO2 at different stages of ALD-growth on (a) PMMA/Si(100) and (b) PVP/Si(100). |
SEM imaging was performed to observe the surface morphology of TiO2 grown on Si(100) and PMMA/Si(100) after 1200 ALD cycles. During spin coating of PMMA, a part of Si substrate was deliberately covered by scotch tape, which was taken off before growth to observe the interface of TiO2/Si and PMMA/Si. Fig. 5 reveals the surface morphology of TiO2 (1200 growth cycles) grown on Si(100) and on the interface of TiO2/Si-PMMA/Si. 1200-cycle TiO2 grown on Si(100) (Fig. 5a) exhibits its grainy surface structure with 5–10 nm sized grains.
![]() | ||
Fig. 5 SEM images of PMMA/Si surface after 1200-cycle TiO2 growth (a) Si(100) substrate surface (b) the interface of Si(100) and PMMA showing the effective inhibition at the PMMA side. |
A boundary (Fig. 5b) is clearly visible at the interface of TiO2/Si and PMMA/Si, where relatively large sized grains are observed at border on Si(100) side and PMMA surface confirms the absence of TiO2 film growth.
Utilization of polymer films for AS-ALD studies brings an extra advantage which is their facile removal after the selective deposition process is completed. PMMA can be easily dissolved in acetone while PVP is soluble in water. After the growth of TiO2 on PMMA with various number of growth cycles, all the samples were rinsed in acetone for 30 seconds followed by XPS measurements (Fig. 6). XPS measurements revealed the presence of O1s, C1s and Si2p peaks with the similar peak intensity from all samples after PMMA removal. Appearance of Si2p peaks from all samples makes it clear that we were successful in dissolving PMMA. It also signifies the importance of utilization of those precursors for AS-ALD processes that do not react with the polymer masking materials. Otherwise, precursors may diffuse into the polymer masking material and consequently making the removal of PMMA much more difficult and even not possible at all. Precursor exposure time is also very critical in avoiding the diffusion of ALD precursors into polymers and reaching the reactive sites on the substrate. In exposure mode (a trademark of Ultratech/CambridgeNanotech Inc.), dynamic vacuum was switched to static vacuum just before the precursor and oxidant pulses, and switched back to dynamic vacuum before the purging periods after waiting for some time, i.e., exposure time. Time scale for precursor diffusion can be decreased by decreasing the pulse length or exposure time of precursor, however, this might result in sub-saturation precursor exposure of the surface leading to less than the optimized growth rate. We have also performed TiO2 growth on PMMA by increasing the exposure time of TDMAT to 40 s and indeed observed film growth of TiO2 on PMMA. In the present case, TDMAT doesn't react with PMMA within the optimized pulse length of TDMAT which makes removal of PMMA with acetone a straightforward job. We also attempted to dissolve PVP in water after TiO2 growth, however PVP was dissolved in water up to 300 growth cycles, whereas PVP removal beyond 300 ALD cycles were not successful.
Based on contact angle, spectroscopic ellipsometer, XPS, and SEM measurements, we confidently conclude that PMMA is the most suitable blocking layer for AS-ALD of TiO2 using TDMAT and H2O as Ti and O precursors, respectively. Hence, we selected PMMA to demonstrate the micron and sub-micron scale patterning of TiO2 using e-beam lithography.
PMMA is by far the most commonly used e-beam lithography resist as it offers nm-scale high resolution, ease of handling, and wide process latitude. Exposure of e-beam to PMMA results in the breakage of its long chain into smaller soluble fragments, which dramatically renders it soluble in a subsequent development step. Utilization of PMMA as a common e-beam resist presents an inherent advantage to use it as a blocking layer for AS-ALD; i.e., it can be patterned to produce nm scale patterns.
E-Beam lithography was performed on PMMA coated Si(100) samples to produce mm, μm, and nm scale patterns of TiO2. Fig. 7a shows the SEM image of post-developed PMMA after exposure to e-beam revealing patterned PMMA free regions of Si. TiO2 was grown on this e-beam exposed PMMA using 750 cycles of ALD growth at 150 °C. Samples were dipped in acetone for 30 seconds, rinsed, and dried, where after they were loaded into the SEM chamber for imaging. Fig. 7b shows the SEM image of patterned TiO2 after removal of PMMA. Growth only occurred at e-beam exposed PMMA free regions of samples and TiO2 lines having diameter of ∼740–750 nm can be clearly observed. Fig. 7c shows the TiO2 lines prepared using the same strategy, however narrower line-widths of ∼150–160 nm were produced. The debris observed between the TiO2 lines in Fig. 7c is most probably the residue left after PMMA removal. Fig. 7d is the SEM image from the interface of the patterned TiO2 and Si(100) revealing the grainy structure of TiO2. Although glass transition temperature of PMMA 950 ebeam resist (95–106 °C) is less then growth temperature of TiO2, patterning of TiO2 is possible because of the high molecular weight of the PMMA used (950 kg mol−1). The higher viscosity of the PMMA prevents reflowing to a certain extent making the patterning of TiO2 possible.31
XPS and EDX elemental line scan was performed to study the linear elemental variation along the TiO2 patterns and presented in Fig. 8. XPS line scan was performed on mm-scale TiO2 patterns due to limitation of X-ray spot size (minimum ∼ 100 μm). A line across an area of interest is selected on the sample and the XPS gathered data periodically along this line. Ti2p intensity was measured in terms of counts per second vs. spatial location along the line and presented in Fig. 8a. A significantly higher intensity of Ti2p peak is only observed at location of TiO2 pattern while intensity at other points was equal to the background (noise-floor) intensity confirming the successful patterning of TiO2. In EDX line-scanning, the electron-beam is aligned to scan across sub-micron scale features and moves along the line at a certain speed depending on the number of data points. The graph (Fig. 8b) reveals a Y axis modulated signal, the Y-height of which is an indication of the number of Ti K X-ray quanta being detected along the scan-line. Clearly, intensity of Ti K X-ray quanta increases only in TiO2 lines which reaffirms the successful patterning of TiO2 line structures. EDX elemental mapping is performed to determine the positions of Ti and O elements at a specific TiO2 patterned area of the sample. X-ray elemental mapping is a useful technique where elements such as Ti and O emitting characteristic X-rays within the inspection area can be indicated by a unique color. After counting the presence of X-ray signal from a specific element, detector places a bright spot of distinct color on the screen indicating the location of that element in an area map. Such an EDX elemental map of Ti and O from a patterned TiO2 area is provided in Fig. 9. Fig. 9a corresponds to the SEM image of patterned TiO2 line features from which elemental maps of Ti and O are collected. Ti K and O K elemental maps are shown in Fig. 9b and c, respectively. It is evident from these elemental maps that Ti and O are only present in the line features which coincide with the TiO2 lines shown in Fig. 9a.
Cross-sectional TEM was applied on TiO2 patterned sample to visualize the area selective deposition. Fig. 10a and b shows the TEM images obtained after PMMA removal from a single TiO2 pattern. Fig. 10a shows different parts of the analyzed area revealing the presence of Si, rectangular pattern of TiO2, Pt, and the area where growth was blocked using PMMA. Fig. 10b illustrates that TiO2 was uniformly deposited on PMMA free area with a thickness of 36.1 nm.
![]() | ||
Fig. 10 TEM image of (a) TiO2 patterned PMMA free region, (b) patterned TiO2 region revealing the thickness uniformity of pattern. |
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra23923d |
This journal is © The Royal Society of Chemistry 2016 |