2015 Volume 4 Issue 1 Pages A0041
In a tutorial paper on the application of free-jet technique for API-MS, John Fenn mentioned that “…for a number of years and a number of reasons, it has been found advantageous in many situations to carry out the ionization process in gas at pressures up to 1000 Torr or more” (Int. J. Mass Spectrom. 200: 459–478, 2000). In fact, the first ESI mass spectrometer constructed by Yamashita and Fenn had a counter-flow curtain gas source at 1050 Torr (ca. 1.4 atm) to sweep away the neutral (J. Phys. Chem. 88: 4451–4459, 1984). For gaseous ionization using electrospray plume, theoretical analysis also shows that “super-atmospheric operation would be more preferable in space-charge-limited situations.”(Int. J. Mass Spectrom. 300: 182–193, 2011). However, electrospray and the corona-based chemical ion source (APCI) in most commercial instrument are basically operated under an atmospheric pressure ambient, perhaps out of the concern of safety, convenience and simplicity in maintenance. Running the ion source at pressure much higher than 1 atm is not so common, but had been done by a number of groups as well as in our laboratory. A brief review on these ion sources will be given in this paper.
The term “high pressure ion sources” and “HP-MS” were used in 1980s to refer to the ion sources (mostly by chemical ionization) with operating pressure at around 1 Torr.1,2) This is not a very high value in our present day standard, but it was much higher than that of EI in those days, and in some cases, differential pumping system was necessary to separate the analyzer/detector from the ion source vacuum. The detection of “macroions” generated from the electrospray in Fenn’s early experiment had led to the rapid development of modern atmospheric pressure ionization mass spectrometry (API-MS).3,4) Atmospheric pressure is in fact, nearly 1000 times the “high pressure” used in the pre-ESI era, and bigger and more powerful pumping systems had become necessary to accommodate this pressure. One important aspect of performing ESI in atmospheric pressure is that the high density ambient gas can provide a heat bath to evaporate the electrosprayed charged droplets and assist the desolvation and desorption of ions. This is in contrast to the electrohydrodynamic spray,5) a high vacuum version of electrospray that can only handle limited solvents with low volatility.
The demand of high sampling efficiency for ion from the atmospheric pressure ambient to high vacuum makes available the modern commercial instruments which are armed with sophisticated multiple-stage differential pumping and ion guides. Besides being soft, ESI is also extremely easy to operate as there is no need to evacuate the source chamber. Although commercial instrument usually comes with an ESI or APCI source housed inside a robust chamber guarded with safety interlock, one can always open it up and turn it into an “open ion source,” or replacing the original sprayer with a custom made ESI or APCI variant with novel functionality.6–11) Such modification does not disturb the vacuum operating pressure, and what needs to be done is to trick the instrument software into believing that the “ion source is safely closed.”
Instead of running the ion source openly, we had sealed the ion source so that the ionization is taken place under a super-atmospheric condition (p>1 atm). We have implemented the super-atmospheric pressure ionization MS with pure electrospray,12–16) corona discharge based chemical ionization,17) and field desorption,18) and had routinely run our mass spectrometer with ion source pressures ranging from several to several tens of atm. For typical analytical application where the solvent composition can be fine-tuned to best-suit the API-ESI, it is unnecessary to pressurize the ion source, though it is possible to get some signal enhancement due to higher gas throughput. For experiment that deals only with aqueous solution or when the heating of liquid sample to >100°C is desirable, super-atmospheric pressure ESI can be a potential alternative to API-ESI. There are several strategies to preserve the high vacuum of the instrument while working with very high pressure ion source.19) Our recent prototype used a booster pump with variable pumping speed added to the first pumping stage of MS to regulate a constant vacuum pressure. Although the optimization procedure, such as the tuning of ESI emitter position inside a sealed high pressure vessel was not as easy as for an API source, the ion transmission was, however, not very much position dependent because all of the carrier gas was driven into the vacuum stage of the mass spectrometer.
The high pressure ESI work was initiated when we were interested to see how the electrospray plume might behave when it is pressed under high pressure. The shape did change, but another benefit (though as expected according to the Paschen law) was that, the electrical discharge that could sometimes disturb the stability and the performance of ESI disappeared. Under that condition, even the electrospray of pure water with not so sharp needle was stable. Coincidentally, at that time, K. Hiraoka (the boss in our group) was thinking of repeating the electrospray experiment under high vacuum using his laser spray method.20,21) The vacuum chamber initially planned for the vacuum electrospray was used instead for our first high pressure electrospray experiment. Under high vacuum, it is not efficient to produce gaseous ion due to the lack of desolvation, but the aqueous based vacuum electrospray (not ESI), has now been developed as an intense projectile beam source for SIMS.22) High pressure electrospray is the extreme opposite of vacuum electrospray, and some of our work on its development will be briefly presented here.
The ion sources used in our studies were mostly consisted of a high pressure vessel made of aluminum alloy, and have an ion outlet capillary that could be attached/detached easily to/from the ion inlet interface of the mass spectrometer. The ion source was pressurized with dry air from air compressor, or 99.995% nitrogen from a high pressure cylinder. In some prototypes, the ion sources chamber were attached with two transparent windows in order to have a better view on the effect of pressure on the Taylor cone and ESI plume using microscope.
First prototypeFor safety and cost saving purposes, our high pressure ion sources were designed to be lightweight and of compact size. But, as mentioned briefly in the previous section, the first high pressure electrospray experiment was performed inside a rather big metallic vacuum chamber (see Fig. 1). In this setup, the emitter and the body of the chamber were grounded, and the high potential was applied to a counter electrode via an electrode feedthrough at the vacuum flange. Like most vacuum chamber, it had two view port flanges and the electrospray condition could be monitored using a long working distance microscope (VHX, Keyence, Osaka, Japan). Although it was difficult to guide the ions out from this bulky chamber to the mass spectrometer, it had allowed us to obtain a good optical observation on the electrospray under high pressure (see Fig. 5b).
The stability for the electrospray of pure water in both positive and negative modes in the first experiment was so good that it prompted us to make a real high pressure ESI source that can be fitted to a MS instrument.12) It so happened that the nitrogen generator in our lab had an active carbon filter (AMF150C-02, SMC, Tokyo, Japan) which is housed inside a cast aluminum chamber with a reinforced glass window. That chamber was just in the right shape for the ion source we had in mind at that time (Fig. 2). So, modification was quickly done by boring a hole on the chamber to fit the Swagelok connector. A Teflon flange was also fabricated to hold the ESI emitter. The ESI emitter was a stainless steel capillary with i.d. of 50 μm from New Objective (Woburn, MA). To provide a more stable liquid supply to the emitter which was held under high pressure, we changed our small and old syringe pump (the one in Fig. 1) with a high linear force syringe pump (PHD 4400, Harvard Apparatus, MA). Charged droplets and ions were transported to the atmospheric pressure ambient via a stainless steel tube with i.d. of 0.8 mm. Two heaters were employed in this ion source. One for the heating of air inside the ion source chamber, and the other one was to heat the ion transport tube for desolvation. Due to the large i.d. of the capillary, the gas flow rate at 4 bar was ca. 17 L/min, so when operated, the high speed gas exhausted from the outlet capillary like an ion spray,23) or sonic spray.24) But the air flow inside the chamber was much slower; therefore, the spraying of liquid was driven electro-statically like an unassisted electrospray.
For API mass spectrometers with differential pumping and ca. 0.5 mm inlet orifice, the typical gas sampling rate is in the order of ca. 1 L/min. That means most of the ion rich gas from our first prototype was not fed directly to the mass spectrometer due to the limitation of gas load allowed for maintaining the operational vacuum. In our second and other higher version prototype, the ion source was connected directly to the mass spectrometer so that all the gas carrying the ion would be introduced to the first pumping stage13) (see Fig. 3).
In the molecular beam research, having a beam source generated from a very high pressure gas reservoir is quite common. One can use a narrow nozzle, or gate the beam source with pulse controlled valve.25) In our second prototype, a modified ion transport capillary with inner diameter of 0.25 mm was used, and this diameter was about half of that of the original capillary. In the laminar flow regime, the gas throughput, Q is governed by the Hagen-Poiseuille equation: Q=πd4(P12−P22)/256ηL, where η is the viscosity and L is the tube length. It is proportional to the forth order of the tube diameter d, and to the second order of the upstream pressure P1 if we ignore the downstream pressure P2, which is much lower than P1 in the case of introducing sample gas to vacuum. For the ion transport tube of 0.25 mm inner diameter, a gas throughput which is equivalent to that of API source with 0.5 mm i.d. inlet could be achieved when the HP ion source was pressurized to ca. 4 bar without additional pumping.13) It was quite a simple and easy coupling strategy but at the expense of ion transmission coefficient.26)
In order to increase ion inlet’s inner diameter, a mechanical booster pump (Roots pump, PMB 003C, ULVAC, Kanagawa, Japan) was later added to the vacuum system and installed in between the mass spectrometer main evacuation port and the original rotary vane pumps.14) Instead of connecting the three phase motor directly to the power line, a variable frequency controller (VF-AS1, Toshiba Schneider Inverter, Japan) was used to adjust the pumping speed of the booster pump, therefore the pressure of the first vacuum chamber could be precisely regulated at a particular value, say 1.95 Torr for any ion source pressure ranging from 1–6 atm.
Ion source for superheated ESIBased on the design of second ESI prototype, a number of super-atmospheric pressure ion sources has been constructed to implement nanoESI,15) probe ESI,15) corona discharged ionization,17) and non-vacuum field desorption.18) Our recent prototype was used in a relatively harsh experimental condition with pressure up to 30 atm and temperature over 200°C.27,28) The purpose of this experiment will be described in the following section. Safety is a primary concern in this case, and the ion source was made smaller with harder duralumin (e.g. A2017) instead of the commonly used aluminum alloy (A5052) in our previous version (Fig. 4). The diameter of the view port was also reduced to 8 mm and the material was changed from glass to quartz. Smaller ion source chamber has smaller heat capacity and thus, consumes less power to maintain a high temperature. If we can forsake the view port, the source can be made even smaller than the one shown here. However, the gap between the emitter and the chamber inner wall should be around 4 mm or larger so that the droplet, built up during the initial or intermediate spraying failure would not short the electrical connection.
We had frequently used the 50 μm i.d. ESI capillary from New Objective for the atmospheric pressure and high pressure ESI at <1 μL/min flow rate. However, for the high temperature (>100°C) experiment, we have not be able to embed this capillary, together with its join and union to the heating element. Instead, a stainless steel capillary (i.d. 100 μm, o.d. 200 μm, and 2 cm long from Misumi, Tokyo, Japan) was used as ESI emitter (Fig. 4c). It was attached to another stainless steel capillary (i.d. 250 μm, o.d. 1/16 inch, GL Sciences, Tokyo, Japan) using a precision crimping tool. The 1/16 inch capillary together with the ESI capillary was inserted to a copper block, and all gaps were filled with thermal grease (silicone and zinc oxide based). The copper block was heated with cartridge heaters and the temperature was monitored with a platinum temperature sensor. The power supply to the temperature controller (OMRON, Kyoto, Japan) was electrically isolated from the ground using an isolation transformer.
The onset voltage, Vonset, for electrospray to take place was independently given by Taylor,29) Smith,30) and Mann.31) For a given emitter geometry and electrode distance, the onset voltage is proportional to the square root of the solution surface tension. The ambient gas pressure has negligible effect on Vonset, and the value for a particular liquid are the same under atmospheric pressure (p=1 atm), super-atmospheric pressure (p>1 atm) or vacuum (p<1 atm). The threshold voltages for inducing the gaseous breakdown and electrical discharge under different gas pressure are, however, different. The voltage to trigger gaseous breakdown is higher under higher gas pressure due to the reduction of mean free path for electrons and this phenomenon is usually referred as Paschen law.32) Except for the case of high vacuum, the discharge takes place more easily under sub-atmospheric pressure condition. This is another reason why the electrospray is preferably conducted under atmospheric pressure rather than sub-ambient even for other non-mass spectrometry application. In contrast, the discharge sources like glow or barrier discharge are more easily stabilized under reduced pressure.33,34)
Due to the high surface tension of water (0.073 N/m), spraying of pure aqueous solution is known to have higher ESI onset voltage compared to organic solvent mixtures (e.g. 50 v/v% methanol/water solution). Thus, even with pneumatic assistance, it is rather difficult to achieve stable and efficient electrospray ionization for aqueous solution because the threshold voltage for the gaseous breakdown could fall below or at about the same value of onset voltage for electrospray. Once the gaseous breakdown occurs, it leads to a corona or arc discharge that seriously affects the electric field near the ESI emitter and degrades the performance of electrospray. Besides using SF6 as sheath gas to prevent the electrical discharge,35) it is also possible to quench the gaseous breakdown completely by pressurizing the ESI ion source with air or N2 to an appropriate level.12)
The effect is shown in Fig. 5 with air as the working. The gaseous breakdown due to the negative corona discharge took place at lower voltage compared to that in the positive ion mode, and that sometimes causes the negative AP-ESI to be less stable than the positive one. Also shown in Fig. 5 are the voltages for the initialization of electrospray and the establishment of different spraying modes for pure water. These voltages were the same for both positive and negative ion mode and were found to be independent to the operating pressure if there was no occurrence of corona or arc discharge. Although there exists an operational region for the electrospray in the positive ion mode, the negative ion mode could only be operated near the threshold of the onset of electrospray, which is of poor ionization efficiency. Increasing the ion source pressure to about 4 bar could nearly eliminate the corona discharge and the electric field on the emitter tip could be raised until the formation of the multi cone-jet mode.
Sensitive ESI-MS of aqueous solution is important in proteomics if it is necessary to preserve the native conformation of the analyte protein during the analysis. Pure electrospray without any assistance is sometimes not up to this task, and nanoESI or pneumatically assisted ESI are usually used. HP-ESI is potentially useful for such application due to the absence of discharge, and if necessary, the emitter potential can be tuned up to the multi-jet mode for signal optimization. As an example shown in Fig. 6, the bovine insulin in aqueous solution could be easily detected with HP-ESI down to the level of 10−9 M with signal to noise ratio, S/N>10. In this case, the ion source is our second prototype with 0.25 mm i.d. For comparison, solution with the same concentration of sample was prepared again using 50 v/v% methanol/water mixture as solvent and was analyzed under atmospheric pressure condition using the original ion transport capillary (i.d.=0.55 mm). The mass spectrum obtained by atmospheric pressure ESI is depicted in Fig. 6b). Same ESI emitter was used in these measurements with the solution flow rate of 1 μL/min.
In sum, even with a narrow ion transport capillary, it has been proven to be useful for the sensitive MS analysis of aqueous sample solution. It is also interesting to note that at low analyte concentration, the mass spectrum obtained from pure aqueous solution appeared to be much cleaner compared to those obtained from organic solvent mixture. This is conceivable because of less surface active contaminants present in the pure water.
Electrospray plume shapeThe shape of the electrospray plume would change drastically from that of 1 atm if it is placed under high vacuum, that is, the electrospray beam becomes narrower when the pressure of the ion source was reduced. This effect, which originated from the space-charge built-up around the emitter tip, is well known for ionic liquid,36) and similar inspection has also been observed for water based solution in Hiraoka’s lab.22) Our earlier objective for HP-ESI was to check if this effect is also true when electrospray is placed under much higher pressure. Figures 7a) and b) show the shape of the electrospray plume observed under different ion source pressures for 2 bar and 6 bar. The electrospray emitter here is a pulled glass capillary from new objective. The onset potential for electrospray in both positive and negative modes was not affected by the gas pressure under the tested pressure range (1–7 atm). The shape the electrospray plume in front of the emitter became more diverged when the ion source pressure was increased. The observation is in agreement with that in vacuum electrospray, though the difference is not as drastic because the ratio of these two pressures was only 3.
Plastic pipette tips had been used by Aksyonov and Williams to conduct rapid electrospray by inserting an electrode wire into the tip.37) Gel loading tip is another plastic pipette tip with a long and narrow capillary commonly used in the gel electrophoresis. Typical inner and outer diameters for the commercial gel loading tip are about 100 μm and 300 μm, respectively. Although it is possible to generate electrospray with these tips, the quality of the electrospray, at best, resembles that of an un-assisted electrospray and the use of organic solvent mixture such as methanol/water or acetonitrile/water is necessary. Its performance under atmospheric pressure should not be expected to be as efficient as a standard or nanoESI source. But the situation changes when the ESI source is pressurized.
Besides solving the electrical discharge problem, another interesting finding with high pressure ion source pressure was that it was possible to achieve a nanoESI equivalent performance using the plastic gel loading tip with aqueous solvent. The ultra-low flow rate electrospray from these tips under high pressure was first noticed by Hiraoka’s student, M. M. Rahman, and we took some tedious step to verify and measure the solution flow rate. The flow rate measurement was conducted by microscopic inspection on the liquid meniscus during the ion signal acquisition, and the comparison with ESI and nanoESI are shown in Fig. 8. Under an optimized condition, a flow rate of ca. 10 nL/min can be easily and reproducibly achieved using plastic gel loading tip with much larger inner diameter (100 μm). Typically, the flow rate of this order can only be achieved with nanoESI emitters of ca. 1 μm inner diameter, and this type of delicate capillaries can be easily plugged by an inexperienced user. The ESI emitter with large i.d. is much easier to handle. Most importantly, even with this low cost emitter, the ionization performance was found to be nanoESI equivalent, i.e., the ion suppression of the hydrophilic analyte by the more hydrophobic compound in the solution is much less compared to high flow rate ESI (Figs. 8c, d). The mechanism to sustain a ultralow flow rate with such a large inner diameter capillary had not been verified yet, but the following factors are believed to be important: i) the hydrophobicity of the emitter capillary and the large surface tension of the aqueous solution make the droplet to stay stably at the emitter tip, ii) the absence of electrical discharge to initiate a stable electrospray for aqueous solution, and iii) possible increase of flow resistance due to high pressure and space charge built-up around the emitter tip.
ESI source is sometimes heated either to aid the desolvation,38,39) to study the temperature effect on the protein structure in solution, and to promote a higher degree of denaturation for protein or DNA.40–42) Most commercial MS instruments also provide the heating element for the pneumatic sheath gas. Recently, a so-called “super-heated jet stream thermal gradient source” was also introduced by Agilent to create a thermal gradient that can focus the sprayed beam at the center axis. Despite its name, the solvent at the emitter tip and the charged droplets are not super-heated above boiling point under the atmospheric pressure. Basically, all kinds of heating employed in the AP-ESI sources are limited to the normal boiling point of liquid, which is 100°C for pure water under atmospheric pressure.
Like the working principle of pressure cooker, the boiling point can be increased by pressurizing the ion source, thus allowing the temperature of the electrospray solution to be further increased beyond its normal boiling point. Figure 9 shows the electrospray of water under different pressures and temperatures. Under 1 atm, the electrospray of pure water was not stable owing to its large surface tension and the occurrence of corona discharge before the onset of stable electrospray. However, the appearance of a water droplet on the ESI emitter before it dropped off from the tip, was clearly visible for a temperature up to 90°C under a liquid flow rate of 4 μL/min. As depicted in Fig. 9a, water droplet disappears due to the complete evaporation when the emitter temperature reached 110°C. A few minutes after the pressurization to 11 atm, the droplet cone re-emerged from the emitter tip (Fig. 9b) and the electrospray plumes right in front of the emitter are clearly visible. Under 11 bar, the boiling point of water was about 184°C,43) and the electrospray was reasonably stable up to 180°C. Although we expected some discrepancy between the measured emitter temperature and the liquid temperature, the microscopic observation here shows that the measured values agreed reasonably well with the boiling points of water.
Superheated water is different from the water at room temperature because it has low surface tension,44) viscosity and dielectric constants,45,46) but has a higher concentration of H3O+ and OH− ions,47) and can dissolve polycyclic compounds.48) Superheated water has already been used in LC chromatography, and the high pressure superheated ESI may provide a tool to further exploit its analytical and industrial potential. Besides the result with water, we had also tested the superheated-electrospray with organic solvent mixture and even neat methanol up to 180°C. At vapor pressure of 27 atm, the reported temperatures for methanol and acetonitrile are ca. 180°C49) and ca. 230°C,50) respectively, thus, in principle, HP-ESI is applicable to most LC eluent.
Rapid in-situ protein digestion MS with superheated ESIIt may seem weird to heat the ESI source beyond 100°C, but superheating the sample solution is what a proteomist would usually do to accelerate the non-enzymatic digestion process with acid hydrolysis.51–55) This type of chemical digestion has a selective cleavage at the aspartate (Asp, D) residue. Microwave oven is usually used for this purpose and ESI-MS is done in an off-line basis. Simply heating the ESI source would turn it into an online digestion reactor. The online operation eliminates the need to transfer the sample from the digestion reactor, and the output of the digestive reaction can be monitored and manipulated by the solution flow rate and heater temperature in a near real-time basis.
An example of superheated-ESI MS is shown in Fig. 10 for bovine ubiquitin under different solution temperatures. When ESI source was heated above room temperature, the protein in the solution denatures, and the unfolding of protein structure was indicated by the shift of major peaks to the higher charge states in the mass spectrum. This thermal denaturation effect was well reported in the literatures. In the case of small protein like ubiquitin, after a drastic shift of charge states at around 80°C, the charge state distribution did not show significant change even when the solution was heated to ca. 100°C. The thermally induced species could be observed when the solution temperature was raised above 140°C. These fragments resemble the y-ion and b-ion in the gas phase dissociation such as that in collision induced dissociation. However, a close verification of the peaks using high resolution mass spectrometer in Fig. 11 shows an abundance of chemical digestion (also called acidic hydrolysis) products. In Fig. 11a, the protein solvent is 2 v/v% formic acid aqueous solution which has recently become a standard reagent for the chemical digestion. Except for the peptide fragment [1–18], all other hydrolysis products are due to the cleavage at the Asp sites. Comparison was also done with acetic acids (Fig. 11b) and the result shows the hydrolysis at exactly the same cleavage sites, but in contrast to formic acid, there is also a significant presence of dehydrated species, e.g. (M+nH−H2O)n+.
In summary, the results here show that the selective acidic hydrolysis (chemical digestion) can be done rapidly, by simply infusing the peptide/protein solution through the superheated electrospray ion source. Both formic acid and acetic acid show the selective Asp-specific cleavage, and a temperature around ca. 180°C was enough to provide the coverage for all Asp sites for ubiquitin and other protein like myoglobin.28) The hydrolysis process was presumably taking place inside the heated stainless steel capillary of the ion source, and the peptide fragments were ionized in-situ when the solution reached the tip of the ESI emitter. The whole processes of digestion, ionization and MS acquisition was done in less than half minute. In addition to hydrolysis, the present method can also be used to study the denaturation of protein or DNA and the analysis of protein complexes in a temperature range that cannot be achieved under atmospheric pressure.
Owing to higher gas density and higher collision frequency for reactant ions and analytes, the performance of gaseous ionization technique such as APCI can potentially be enhanced by increasing the ion source pressure. Electrical discharge is usually used to create reactant species for gaseous ionization, but higher voltage is needed to initiate and sustain a stable discharge plasma compared to those under vacuum or 1 atm. Nevertheless, high pressure discharge beyond 10 bar had long been tested,56) and has been used for laser or molecular beam source excitation.57) In some extreme experiment, it is even tested up to the supercritical points of CO2.58)
Although there has been little report on the super-atmospheric pressure chemical ionization mass spectrometry, there do exist some earlier reports that show the enhancement of gas phase ionization efficiency under higher operation pressure. One example is the high pressure ion mobility spectrometry (HP-IMS) in Hill’s group.59) The high pressure buffer gas was originally used in an attempt to enhance the resolution of IMS, and the ion source was radioactive isotope 63Ni. Besides the resolution enhancement, maximum ion signal was obtained at an optimum pressure above 1 atm. (We got similar result using high pressure corona discharge ion source.17)) Another example is a high pressure gas phase femto-second laser ionization source by Peng et al.60) Helium was used in the experiment, and a better performance was obtained with helium carrier gas pressurized to >3 atm. Helium was difficult to pump away, and small ion inlet aperture had to be used in their experiment. Nevertheless, this problem can potentially be solved using gated valve.
A brief review has been given to the super-atmospheric pressure ionization sources, mostly on ESI done in our laboratory in the past several years. It might seem odd to pressurize the ESI source at the first place, but outside the field of mass spectrometry, high pressure electrospray has long been tested,61) and used in electrospinning,62) and electrodispersion of water in CO2.63) In contrast to high pressure discharge, stable electrospray is relatively easier to initiate regardless of pressure, and the gas phase ionization by electrospray plume is also predicted to be better under super-atmospheric condition.64) Thus, one interesting future work would be to use the high pressure secondary ESI for the gas phase ionization of gaseous compound. The main difficulty to incorporate a high pressure ion source to the mass spectrometer is the ion sampling interface and further development and innovation on this aspect are still in need.
Most of the works in this paper were supported by the Program to Disseminate Tenure Tracking System from the Ministry of Education, Culture, Sports, Science and Technology of Japan and the Grants-in-Aid for Scientific Research from JSPS.
