| Author | Message | ||
     Michel Coene (michel) Advanced Member Username: michel Post Number: 34 Registered: 2-2002 |
I did an application where we measured active ingredient in glass vials. A tip: if you use polarised light, you can set the direction of polarisation horizontally, and this has the effect that you can measure on the side of the vial and the light does not bounce off the glass if you don't hit the exact center of the vial. You should certainly consider this advantage if you plan on scaling up to online. AOTF crystals produce polarised NIR light (when used without fibers). I am sure Gaby Levin will be able to tell you more about that. You might however also just work with a halogen lamp, a polarisation filter and a diode array... | ||
| Jonathan Hoggatt (jhoggatt) Junior Member Username: jhoggatt Post Number: 7 Registered: 6-2006 |
The cake size is fairly thick (25-30mm) and the product is hygroscopic. Shaking has been considered, but you have to basically sonicate these cakes to break them apart without opening the vials. Also, we are hoping to scale this project up, so shaking of the vials in the long run is not very feasible. I had considered using a spinner or a fiber optic probe to sample from the side, or a combination of both... | ||
| Michel Coene (michel) Intermediate Member Username: michel Post Number: 33 Registered: 2-2002 |
If this is a lab assay, can you not just shake the vial before placing it on the analyser? | ||
| Yiwu Zheng (eric_yiwu_zheng) New member Username: eric_yiwu_zheng Post Number: 1 Registered: 6-2006 |
Hi Jonathan, We have measured residual moisture content in freeze-dry protein products before. The questions you asked are dependent on how much water contains in your sample and how thick the cake is. If the pathlength (sample thick) is not long, the NIR radiation can pass through sample cake. The gradient of moisture content may not be a real problem. We recorded 4 spectra for each vial at four different angles. Before calibration, EMSC and/or EMSC+2nd derivative were applied to separate physical influence. You also may sampling using a spinner. However, it did not help much for my samples. Good luck, Eric | ||
| Gavriel Levin (levin) New member Username: levin Post Number: 3 Registered: 1-2006 |
Hi Jonathan, Lars Sukowski from Roche Basel did excellent work on moisture measurement. It is currently being done there at high speed of 300 vials per minute. I can not talk too much, please contact me so I can send to you the relevant chapters from a Ph.D. dissertation work done by him on moisture determination in lyophilized vials. Gabi Levin Brimrose Corp. [email protected] | ||
| Jonathan Hoggatt (jhoggatt) Junior Member Username: jhoggatt Post Number: 6 Registered: 6-2006 |
I was wondering if anyone has experience with moisture determination of lyophilized products using an integrating sphere. Specifically, I am curious about cake uniformity. If I sample my product by placing the vial on the sphere and sampling from the bottom, my calibration samples and unknowns for that matter will all be for the bottom couple of millimeters of the cake. However, in order to do a reference method (such as Karl Fischer) I would have to break up that cake and essentially sample a mixture of the entire cake. My question is... what if the cake does not have a uniform moisture content? For instance, if there is a leak in the product vial, it is possible that the moisture content would not be uniform throughout the cake, but rather would be a gradient of sorts... highest moisture at the top of the cake, due to exposure to ambient air (read: humid) and lowest moisture at the bottom of the cake due to a lack of exposure to said air. Since the NIR sampling was from the bottom of my vial, I will be getting values that are lower than the actual "total" moisture content of the entire cake. It seems like the correlation between the Karl Fischer values and the NIR spectra would be affected. Does anyone have experience sampling on these types of samples, and if so, is this variation in cake uniformity really that big of a concern in developing an accurate and robust model, or am I just over-complicating things? If it is a concern, what sampling methods are commonly used to avoid this? |