I have a general question about column selection, in particular the effects of column ID. Does column ID affect the lower detection limit of a system? I know column ID affects column capacity but I'm trying to rationalize if that affects detection limit. For example, if you want to detect a 1ppm sample with a 1uL injection, that's 1 nanogram on the column. Will I get a higher area count for this peak if I use a 0.10mmID column or a 0.53mmID column or does it not matter?
By Anonymous on Friday, March 7, 2003 - 12:02 pm:
You should have a better detection limit with the narrower column, but if you optimize the method you use for each column you will have a significantly lower area count for the narrow column. This is because detection limit for a given detector is not based on area but on signal to noise ratio. You usually see detection limits expressed as mass per unit time to put the detector sensitivity on a column independant basis. For example, if a peak on a 0.10 mm column elutes in .5 sec and the same peak on a 0.53 mm column elutes in 10 sec it will take 20 times as much compound to see the compound on the 0.53 as the 0.10.
For the 0.53 column the analysis will probably be done by splitless injection. To optimize the column loading in the 0.10 column a split injection will have to be used. Since the peak is narrower on the 0.10 mm column the signal to noise and the detection limit expressed as minimum mass detected (not per unit time as expressed above) will be lower than the minimum mass detected on the megabore column even though the area counts on the 0.53 will be much higher.
By Leon on Thursday, March 13, 2003 - 07:03 pm:
A column ID is not the only parameter that affects the detection limit of a GC System. Column dimensions (length, ID, film thickness) and the method of analysis – all affect the detection limit. On top of that, the very concept of a detection limit depends on the experimental constraints.
It is convenient for the simplicity and the clarity of the answers to assume that all (open tubular) columns in the following analysis have the same plate number, the same type of a stationary phase, and the same phase ratio. This means that a column with smaller ID is proportionally shorter, and has proportionally thinner stationary phase film. I also assume that each injection is sharp enough for its column, and the temperature program (if any) in the method utilizing smaller column is appropriately faster than the one utilizing larger column. (This is known as Method Translation). In addition to that, certain assumptions have to be made about the detector electronics and the data analysis system. Without going to more details, I assume that the whole system is capable of “similar” handling of the sharp as well as the wide peaks.
Suppose now that (does not matter how) equal, NON-OVERLOADING AMOUNTS of a pure compound were injected in two columns having different IDs. As a result, although the peak coming from the smaller column will be narrower and proportionally higher than the one coming from the larger column, both peaks will have EQUAL AREA COUNTS. The key to the detection limit in this case is the electronic noise in the detector and the data analysis system. Typically, for the peaks with equal area, the narrower is the peak the smaller are the relative errors caused by the noise. As a result, in order to have the same relative errors, a narrower peak can have smaller area.
The verdict. For the appropriately scaled conditions, the smaller is the column ID the LOWER (better) is the Minimal Detectable Amount (AMD).
It is well known, however, that, one of the major problems of fast GC utilizing narrow bore columns is its higher (worse) than conventional detection limits. The reason is also known. The sample capacity of a column (maximal amount of a sample that can be injected without overloading the column) is proportional to the third power of ID (reduction of ID by a factor of two reduces the sample capacity by a factor of eight). A typical sample has more than one pure compound. One might be concerned not only with the detection of the smallest peak, but also with making sure that the high concentration compounds do not overload the column. If that is the case then one can no longer presume that equal amounts of the same compound can be injected in a column regardless of its ID. Reasonable test conditions for this scenario could call for maintaining of a relatively equal overloading in all columns. This can be provided by injecting equal, NON-OVERLOADING CONCENTRATIONS of a sample in all columns. Unfortunately, this means that the smaller is the column ID the smaller is the NON-OVERLOADING AMOUNT of the sample. (two-fold reduction in a column ID causes eight-fold reduction in the non-overloading sample amount). This is unfavorable for the detection limit.
Different verdict. For the appropriately scaled conditions, the smaller is the column ID the HIGHER (worse) is the Minimal Detectable Concentration (AMC).
A possible method development strategy. To speed up an analysis, use the column with the smallest possible ID. If the detection of the smallest peaks is a problem, inject more sample. If the large peaks become overloaded beyond an acceptable level, consider increasing the column ID (or other tricks, such as another detector, a better data analysis, etc.).
By Leon on Thursday, March 13, 2003 - 07:19 pm:
Of course, Minimal Detectable Concentration should be abbreviated as MDC (not AMC as in my previous comments).
By Anonymous on Tuesday, March 18, 2003 - 09:18 am:
I can follow the logic and agree that for equal amounts on column the area should be the same in theory, and that a narrower column will have a lower minimum detectable mass. I don't follow the logic for the lower maximum loading increasing the detection limit and making the detection limits worse for narrow bore columns. The limitation on loading affects the dynamic range, not detectability.
There is an issue with the concentration in solution that can be detected on a microbore column due to the high split ratios required, so that low concentrations cannot be run splitless. If you look at the literature you do not see minimum detectable concentrations in most cases, you see detection limits listed in mass per unit time. This is an attempt to make the detector sensitivity nujmbers independant of column diameter.
If you inject the 1 ppm sample in the first posting on the two types of columns listed using the appropriate conditions for each column, you will have a signal to noise ratio for the narrow bore column at least as good as that of the wide bore column, thus the detection limit will be as good or better. The dynamic range will of course be much lower for the microbore column, so I would expect problems on the high end, not on the low end.
By Anonymous on Tuesday, March 18, 2003 - 12:39 pm:
Real world analyses do not always follow theoretical assumptions. Microbore columns require operator expertise and state of the art data handling systems to collect usable data.
Inherent factors in performing an analysis dictate the optimal choice of column ID and film thickness. Are we interested in trace analysis, bulk purity, clean samples, dirty samples, air monitoring, gasoline analysis, PCBs, pesticides, blood or urine analyses?
Different analyses require different columns, much as different requirements of vehicular transporation dictate a truck, a minivan, or a subcompact automobile to be the optimal choice to get the job done.
There is no simplistic choice and the chromatographer must use his experience and knowledge to make a better if not a best choice(compromise) to get the job done quickly and cost effectively.
I have used microbore columms which tolerated only a few hours of dirty samples and megabores which have lasted for months and the difference in analysis time was only 2-3 min per analysis.
The absolute most definite answer is..... it all depends.
By Leon on Thursday, April 3, 2003 - 12:29 pm:
To Anonymous on March 18, 2003 - 09:18 am
1. The original question (raised by Anonymous on Thursday, March 6, 2003) was: “Does column ID affect the lower detection limit of a SYSTEM?”. It is not about the detection limit of a detector (which can be measured without a column), but about the detection limit of a GC SYSTEM (that includes a column). And you wrote, “narrower COLUMN will have a lower minimum detectable mass”. Let’s continue keeping our focus on the key subject – the effect of the COLUMN ID on the detection limit of a GC SYSTEM.
2. For a system, the “detection limit” is a generic term representing several metrics related to ability of the system to detect and to quantify the trace level components. Which detection limit metric should one use depends on the analysis goal.
3. We both agree that (I quote you again) “narrower column will have a lower minimum detectable MASS”. One might rephrase it by saying that narrower column is more favorable for detecting fewer picograms, femtograms, etc. of a trace component.
4. In majority of real analyses, one is concerned not with the picograms, femtograms, etc. (because much more than that amount of a trace component is typically available in a sample vial), but with ppm (parts per million), ppb (parts per billion), etc. concentrations of the trace components. In these analyses, (and I rephrase you again) wider “column will have a lower minimum detectable” CONCENTRATION. One might rephrase it by saying that wider column is more favorable for detecting fewer ppm, ppb, etc. of a trace component. Why is that? Because, as you pointed out, narrower columns can tolerate lower undistorted sample loading. You say that you “would expect problems on the high end, not on the low end”. I agree, and that is exactly where the detection limit problems begin. In order to compare apples with apples, that is to have the same resolution of all peak pairs (including the high concentration peaks) in both cases, one needs the same distortion level of the high concentration peaks in all cases. For that, one needs to reduce the amount of the sample injected in the narrower column. That is precisely why narrower columns require higher split ratios that you mentioned. As a result, if, as you suggested, you inject 1 ppm sample in all cases, the MASS of the sample injected in the narrower column will be much smaller, and the trace level components that might have been detectable in a wider column might become undetectable in the narrower column. This is not only a logical conclusion, but also a well known experimental fact. Typically, MDC increases in inverse proportional with square of a column ID or even sharper (additional details are necessary for more accurate predictions).
To Anonymous on March 18, 2003 - 12:39 pm
I agree with everything you write. Indeed, the task of a GC method developer might be very complex and unpredictable. The question at hand, however, is clear, specific and simple. “Does column ID affect the lower detection limit of a system?” A simple question deserves a simple answer if there is one. Luckily, this is the case, and “the absolute most definite answer is..... it all depends” on the column ID in exactly the way predicted by theory and confirmed by many experiments, as I described earlier. You can check it for yourself if you doubt.
You also mentioned that in your experiments with the microbore and the megabore columns, you observed that, in the first case, “ the difference in analysis time was only 2-3 min per analysis”. Certainly, one can design a microbore-based method that can be only slightly faster than the megabore-based one. If necessary, the former can be designed to be even slower than the latter. It all depends on one’s goals. Theory predicts, however, that, when one of the goals is to preserve the resolution of all peak pairs, the analysis time CAN BE reduced in proportion with the reduction in the column ID or ID squared (more details are needed for a specific prediction). This has been verified in a large number of method translation experiments made by many chromatographers, and I am yet to see it failing to behave exactly as predicted by the theory.