It seems like you need somebody whom is ceramic expert. However, the explanaiton you wrote suggests some specific weight change, which is translated in different ways upon the software you are using. Alternatively, a magnetic/paramagnetic behavior is also possible, depending on the ceramic raw materials you have.

The crucible were not contaminated? Do you subtract the base line?

At what temperature does the weight loss/gain occur? You might need to check if it is the moisture in your sample that happens around 100 C and keep it in a dry box for 48h. Otherwise, you can also try pure inert gas to shut down the oxygen effects if it happens to Ferrite. Then it might give you another evidence to rule out the possible effects.

Good Luck !

Hi, the results look puzzling effectively.

Normally, you have two pans, one receiving the sample, the other one being the reference. Are they made both of platinum in both instruments?

What about the Archimede correction with temperature? Can you anticipate that the local geometry ans space within the systems are different enough?

Sorry, at the time, more questions than answers

I think that you have to repeat your samples synthesis or to take one of instruments and based on its results

I dont know the explanation.This is an experimental question. You can change the pan by one inert. The platinum can catalyze a reaction in one of the two systems.

Note: I have the Mettler TGA/DTG system

Sometimes with porous solids it can be useful to press them inside the crucible, to avoid buoyancy effects that can give thermograms with a weigth gain, quite unusual in keramics. I had a sepiolite sample that every ti me was giving me a weigth gain, just pressing it in the pan and it was ok. Good luck man!

Have you tried with other known substances? Is it only your substance, that shows this result?

Are you sure that the sample pan and the reference pan were placed correctly? In the case of Diamond TG/DTA, the LHS is for the reference and the RHS is for the sample.

Is the DTA signal (endo/exo) reversed too?

This system should neither lose or gain weight (maybe a small loss of adsrobed water which would me more of a drift in the base line than a TGA stair step. You have a magnetic effect! Depending upon polarity of the furnace it will either be gain or loss. Gheck the Niel (spelling is wrong) Temperature of the material.

To state anything about your Problem, it would be helpful if you could send the different graphs of you Investigation, the experimental conditions as well as the respective background (bouyancy) curves.

Did the samples come from the same batch of material? Is it possible that you have impurities in one that do not exist in the other?

some interesting information in this M.Sc. thesis...

Do you get any response from the heat signal at all? Is your weight changes phases related? This may be relevant especially as your using a TG/DTA system on one hand and a TG/DSC on the other. Absorbed water should definitely be considered, however your best way of finding out is to heat up your system inside the TG furnace with a heating cycle then cool off and redo the analysis right after. You will have some weight differential to work out as your starting weight on the second run may not be at 100% but if would eliminate any issues related to undried samples.

Something else, did you use the same quantities of weight in both instrument? Are the instruments set on the same gain and are you within their balance range concerning accuracy?

Some of TG equipments have an additional magnetic field for calibration of temperature using Curie points.

Try using alpha-alumina as reference material, then a blank run under same conditions of sample and substract base line from sample run.

Hi, can you send a copy of your results to better understand what is going on. What temperature did you observe the onset of weight changes. If the results were consistent throughout the three samples, then the most likely culprit will be the inherent error of each system, Also check to make sure sample pan was carefully tared and the electronic scale is properly calibrated in each system.

The best thing is to try standard sample with known phase change behaviors and see if there is a mass loss or gain.

A third piece of test equipment would settle the score! Try it on a TA Instruments piece of equipment.

In all seriousness, have you run the experiment in Nitrogen? Are you using ultra high purity air? Do you have a dessicant dryer in line?

Assuming that you have done all the calibrations correct etc checking with known materials that give exact mass loss changes on both machines then there are a few things to check:

If you material is sensitive to gas flow i.e. is the gas flow experienced at the sample the same in both machines (are the pan diameters, heights etc the same?). The material you are working with is sensitive to forming impurity phases such as Bi25FeO39 and Bi2Fe4O9 which doping with Ca+2 changes the Néel temperature as noted in a previous comment. Another point to consider.

What is the “DSC” signal telling you in these weight changes?

Another thing to think about would be to run your samples at different heating rates and see what affect that has – often it can reveal quite a bit about what’s going on.

And keep positive and trying!

I also believe you must try a standard sample with a known thermal behavior as proposed by Mr. Shehdeh Jodeh, because this may be helpful in analysing the quality of your equipment. It most be noted that weight loss or gain for this sample is quite surprising and only desorption or adsorption may justified such behaviours.

there are several things that can affect the TGA signal.

first: baseline. You say that you have cheked the baseline but typically user check the baseline in empty pans. Please, use a inert material like a igh temperature ceramic with a sample mass of the same value of your sample. this experiment will show if the discrepance between the machines came from the instrument itself.

second: oxidation of the sample. the oxidation is a proccess with weight gain. you say that you are using air. this means that probably this is not the problem but please the the quality and the purity of the air. misture could also cause inconsistencies. are you using the same sample amount in both machines? are you using the same type of pans? some pans let the sample more expose to the atmosphere than others.

thirth: pans. Are you using the same pan material?. Typically there are 2 types of pans, ceramic and platinum. But the platinum pans in some providers are not 100 % platinum but have some rhodio. perhaps there is some reaction between your sample and the pan. I agree with all the comments made, with the original data and the experimental conditions would be easier

These may be obvious questions but you do not say what sample mass you are using and whether there is any difference in the size of the Pt pans used or the purge rate? Occasionally I have seen differences in thermal behaviour with different instrument geometries esp when there has been insufficient purging. What I would suggest, if it is not too obvious, is to try re-running the samples of both machines using a sample size of ~ 5 mg with a purge rate of 100 ml per min and see if you still see any differences in their thermal behaviour

The two different system that you talked about, are the chamber of both system made of same material? The difference may be caused by the type of coating inside the heating chamber. Since you are dealing with ceramic doped with metal, you might want to check if there were any reaction where weight gain is applicable.

A weight gain in TGA is unusual (except when a chemical reaction takes place). Were both the experiments conducted in similar conditions (atmosphere, heating rate etc)? Also try to convert them into % weight loss vs temperature thermogram.

In the initial years of Thermo balance design, the pan and furnace were at the bottom whereas the quartz rod (hanging the balance) which acted as the core of the LVDT was at the top. They found some weight gain due to the deposition of vapour emanating from the sample. In order to overcome this problem, Chevenard redesigned by reversing the two systems (keeping the pan and furnace on top) so that vapours go out without deposition. So weight is not expected with temperature when thermal degradation takes place, especially when an inert atmosphere is maintained around the sample.

I agree with Carlos, The air atmosphere maybe leading to oxidative weight gain. Moreover your total weight loss you mention is around 0.5%. Please ensure the method followed for both the instruments. Or else create a method with heating rate and final temperature only.

I do not know the Perkin Elmer system in detail, but fairly good the NETZSCH. Nevertheless, I guess that the problems might arise from interaction between sample and gas. Even if you write that the runs are performed in air, at least the NETZSCH typically uses additionally a "protective gas" flow (usually Ar). This dilutes the oxygen concentration of the overall gas. Maybe the same happens with the Perkin Elmer, but with different flow rates? And the ferrites are for sure very sensitive against oxygen partial pressure and temperature, because valency easily changes Fe2+ Fe3+ (FeO, Fe3O4, Fe2O3, see J. Crystal Growth 275 [2005] e849–e854, available here on Research Gate). But before discussing these redox equilibria in detail you should again crosscheck the behavior of both equipments with an inert sample, such as Al2O3.

You carried out the experiment in air atmosphere. The weight loss is is quite expected because of decomposition of the sample. The weight gain is not expected ...but in your case, as the heating is done is an atmosphere of air there is a possibility of oxidation to higher oxides of one or the other metal leading to weight increase. But this need not be dependent on the instrument...especially at higher temperatures.

The first thing I would check is difference in atmosphere between the two instruments. The other thing to look for is whether the measurements were conducted in static or flowing atmosphere. I have seen differences similar to these in these cases, usually associated with gas adsorption/desorption.

The unusual weight gain/loss could bet from the apparent mass change resulting from the change in the density of the gas around the sample on heating .In the case of bottom loading type of thermobalances one expects an apparent mass gain due to decrease in the density of the gas on heating. This apparent contribution to the mass change also depends on the geometry of the sample crucible since the convection in the reaction chamber can exert pressure on it when the cold gas descends into the hot zone where the sample pan is located.. such apparent mass change at times can mass the real mass change in the material when the total mas change in the sample due to any chemical reaction is small. In the top loading thermobalances the effect may be opposite to that of what is observed in bottom loading balances. In this case ( as in the case of NETZSCH STA top loading unit the cold gases cannot rise against gravity and hence will not exert any thrust on the sample holder. The density of the gas around the sample will decrease and hence result in the apparent change. Such apparent changes have to be carefully evaluated in the case of adsorption studies where the pressure in the system is continuously increased to follow the extent of adsorption as a function of the pressure of the adsorbate gas. Such types of errors could also be observed when the density of the gas around the sample is changed due to the change in the composition of the gas mixture, as in the situation where the mixtures of CO and CO2 gases are used to obtain different partial pressures of oxygen in the reaction zone.

It seems important to check for any possible unwanted magnetic interference in the thermogravimetric essays, being the sample ferromagnetic. Nearby strong permanent magnets or electromagnets located below the height level of the sample may cause an apparent weight gain ─ an experimental artifact ─ below the temperature of the ferromagnetic Curie transition, when ferromagnetic samples are analysed. If located above the height level of the sample, they may cause an apparent weight loss. Such interference can eventually be caused by solenoid valves, among other possible causes. Working with ferromagnetic samples it would be convenient to check for this possible cause of experimental artifacts, what can be done by carefully looking for possible causes of magnetic interference and checking for the absence of a gravimetric signature for the ferromagnetic Curie transition. If such a signature is found, being the effect reversible, the cause of magnetic interference must be removed or at least identified so that the experimental data can be interpreted accordingly.

Dear Friends,

First let me express my gratitude to all of you who have share their invaluable experience and expertise with me in understanding the problem at hand. I am highly obliged to all of you.

I would also like to convey my sincere apologies for not responding to your queries / suggestions earlier. The delay was caused because we are running several experiments in our lab, which were based on the suggestions that I have received from all of you. What I have understood from these experiments is that the design of the balance (Top or horizontal) and environment (amount of flow) plays major role as bouncy effects are different in both these circumstances.

As the change is mass, which I need to measure is very small, what I found is precise cancellation of background signal is very very important. To understand this, we put Al2O3 of equal mass as reference and sample, measured its base line first and then in the sample pan Al2O3 was replaced by sample. In that case I found similar trend of data obtained on Netzsch and Perkin system.

But still I am not satisfied with the result. We have make measurements both in the heating and cooling cycling from RT – 800 deg C in Air atmosphere with a heating rate of 5 K/min in Pt-Rh crucibles. During heating cycle, it shows a weight loss (due to loss of water of crystallization) till 350 deg C followed by gain due to oxidation (see attachment). This is expected. But during the cooling cycle, it continues to accumulate further weight, which in my opinion not possible or should not have occurred.

I feel that still base line is not fully compensated for cooling cycle. The base line trend and sample trend during the cooling cycle is exactly the same while such was not the case during the heating cycle. I want your suggestions / guidance on the following aspect, which I feel that would be help me in understanding this mysterious results.

Under what circumstances I would obtain an ideal TG base line. Especially during the cooling cycle do I need to take any special care or need to make any changes in the experimental conditions ?

Note: Please see the attachment for TG and DSC curves.

Once again thanks a lot and with kind regards.

Sincerely,

Bhupendra

In continuation with last conversation.

Dear Dr. Chudasama,

The way you do the tests is fine. Anyway, in the TG-plot the baselines for heating and coóling are mixed up. To improve the data there is one thing I would recommend. Before you carry out the baseline runs please heat the sensor with the crucibles to 1000°C. This would make sure that everything is clean. Carry out the baseline 2 or 3 times to make sure that everything is fine with the instrument. If you want to have best accuracy for the TG-result, please use as much sample material as you can get into the crucible. If you only look for the the TG-result, you can even fill the reference crucible. The DSC signal would be useless but you can double the sample mass and therefore the accuracy on the TG-result easily. If it is crucial for the research to get best possible accuracy you might think about using a pure TG-sample carrier if available in your lab. This would allow you to use initial sample masses into the gramm-range. and the level of uncertainty gets below 0.01%.

By the way, if there is a stoichiometric change of the sample by oxidizing atmosphere, you might carry out tests in air and nitrogen. In nitrogen you should no see any effects and no weight gain.

Hi Dr Blumm Could you please clarify your statement re placing sample in the ref pan. I am unclear as to how you would determine the weight change if you fill both pans. Thanks

Dear Dr. Blumm

Thanks for the detailed response. The usual practices and precautions which you have suggested for recording the baseline are normally being followed in our lab. The difficultly that we are facing is we are trying to measure a very small change in mass ( < 1 %) over the entire heating range (normally 25-1000 deg C). Now, the baseline signal is almost matching the sample signal and any fluctuation in the background while sample measurement or in the baseline data itself gives us incorrect nature of weight loss or gain. I am not aware of the practice which you are suggesting to improve TG accuracy by using samples in both the pans !! Can you please elaborate how actually the microbalance measures mass change if both sample and reference are filled with the sample itself.

Thanks and with kind regards.

Sincerely,

Bhupendra

The microbalance in an STA 449 measure the weight change of the sample holder and everything what is put on it. The sample holder and the crucibles are made of inert materials. Therefore, the only thing which can change weight is the sample in the sample crucible druing a usual test. The reference crucible is generally kept empty during a TGA-DSC test run. In case you fill the reference crucible with material, as well, there will be no proper DSC signal any more. As the balance is connected via the sample holder rode to the DSC sensor and both crucibles, it measures the weight change of the sample and the reference crucible content. Therefore, it is possible it increase the sample weight for a TGA measurement without changing the setup. Please make sure that you put the total sample mass (sample and reference crucible) in the software. Otherwise the calculation of the relative mass change gets wrong.

I have been in thermogravimetry for several decades now, I extensively investigated instabilities in TG's, and from these experiences, I am supporting the explanation given here by Dr. S.R.Dharwadkar R.avalnath Dharwadkar . The details of my experiences are:

Of the two main sources of weiight  disturbances in a vertical TG: buoyancy (floatation, static) and drag (resistance to flow, dynamic) - buoyancy is the stronger of the two, most of the time, except for the very start of the run. There are three kinds of flow in a vertical TG, more strongly pronounced in the furnace-below-the balance variety: (a) a general flow imposed by the flow controllers; (b) “bubbles” of hotter and cooler gas, moving convectionally in an erratic manner; (c) toroidal circulation taking place when the furnace begins to deliver more heating power than before, especially at the start of a TG run. When the furnace starts heating, the walls become hotter than the center of the reaction tube. A strong vertical circulation starts, upward at the walls and downward in the center. The sample holder is exposed to a downward drag force, producing a sharp apparent weight gain. As the over-all temperature increases, the velocity of the circulating flow diminishes, and so the drag force subsides. However, during temperature ramping, some temperature differences between the furnace walls and the sample and generally the tube's center remain, which produces bubbles, or pockets, of different temperature ->density ->buoyancy of the gas. Shielding the sample can protect the sample from a drag, but not from density/buoyancy changes. These dynamic disturbances are stronger during ramping than during isotherming, but buoyancy-related deviations are inescapable even at a constant temperature - unless under vacuum. The dynamic disturbances depend on the volume and the shape, buoyancy on the volume only. Using the sample's mass for buoyancy corrections can help, but is inadequate.

Such deviations are much weaker in horizontal TG's, but that configuration has its own errors, very grave indeed.

http://goo.gl/BjmhRd

It is good that you are showing the TG curves with those typical disturbances. No mystery! All those problems have been resolved and solutions are provided in the article "Precision thermogravimetry", J Therm Anal Calorim (2015) 120:139-147

DOI 10.1007/s10973-014-4384-0   Sorry for promoting my own paper.

Honestly, I didn't read all the replies to your issue, so I don't really know, that meantime you've got an adequate answer to your question, but beside the possible sources of error already listed by others (buoyancy, etc., etc.) the problem also could be the magnetic interaction between your sample and and the magnetic force generated during heating of the furnace. You wrote, that you measured bismuth ferrite, that is a magnetic ceramic material. Magnetic materials can interact with magnetic forces, and this interaction depends on the construction of the apparatus, of the furnace, on the balance etc, and obviously this interaction can be attraction, or repulsion, observable as weight gain, or wight loss. This interaction disappears when you reach the curie point (Tc) of your material, but some ferrites can have very high Curie points, thus the interaction is present during the whole measurement. Actually, in this case you're doing thermogravimetry in magnetic field (TGM). I have quite long experience of the thermogravimetric measurement of magnetic materials, with various instrument types, geometries, etc.

It may not be possible to optimize both the TGA and DTA conditions in a simultaneous TGA-DTA run; one of them needs to be sacrificed for the benefit of the other. TG measurements are quantitative, and the purpose of DTA is usually merely orientation what processes are taking place at which temperature, then optimization of the conditions solely for the TGA seems justifiable. In a TGA experiment the main culprit of the disturbances are physical interactions of the sample's volume with the gas (buoyancy and drag), then equal volumes - NOT masses - of the sample and the reference - count. As always, subtructing the blank run is a powerful corrective tool. I recommend reading my article "Precision Thermogravimetry", cited in my earlier entry in this discussion