[FAQ] Troubleshooting MELTS_Excel e.g. 'Server returned no results, giving up!'

[Paula]

Caveat: I have not used MELTS for Excel very much, but Paul Asimow and I have demoed it at a couple of workshops and helped a few people to get it running at Caltech. I do not have access to the server (although I am somewhat familiar with the MELTS code it uses), and have only explored the Visual Basic code to the extent necessary to get the "Combine tbl" tool to work on MacOS. Any inaccuracies are mine. -Paula

We often get messages like this, either by email or via Facebook:

Excel gives the error "Server returned no results, giving up!"

The error message that MELTS for Excel gives is not very helpful, as it does not distinguish between a connectivity problem and the calculation failing for some reason. It would also be easier for us to debug if we were given a bit more information by the user e.g. have they successfully run MELTS_Excel previously, or has it never worked? What kind of calculation was he/she doing and on what kind of composition? Here are some ideas for troubleshooting MELTS_Excel, and for making it easier for us to help you if it still doesn't work.

First, make sure that you have the latest version of the MELTS_Excel file; April 24th, 2018 the last time I checked. This version does work in Excel 2016. It also implements the mixed H2O-CO2 fluid model (i.e. rhyolite-MELTS 1.0.2, 1.1.0 and 1.2.0 are available), as well as the pMELTS model. When you download the MELTS_Excel file make sure to keep a copy with the default composition and settings (Early Bishop Tuff, v1.0.2 etc.).

In Excel's default documents folder, probably 'Documents', there will be a file called MELTS_Excel.log. If you experience problems, check the contents of this file (if you double-click it should open in Notepad). If it is not obvious what the problem is then open the original Early Bishop Tuff version of MELTS_Excel and try "Find Wet Liquidus". If that fails, then go to the sections on Internet connection / Server problems. Note that it's a good idea to close MELTS_Excel.log between individual calculations.

Calculation problems

If the EBT calculation succeeds, but your own one doesn't, then it might help to work through the MELTS_Excel version of the Quick start using MELTS tutorial. The version of MELTS_Excel used there is older e.g. it is missing the "Find Wet Liquidus" and rhyolite-MELTS/pMELTS model choices. But none of these changes affect the simple tutorial example. Some things to think about in your own calculations:

• If you use "Find Liquidus", "Find Wet Liquidus" or "Equilibrate" on the "Input" tab, the P-T conditions are not automatically transferred to the Sequences tab. You need to re-enter the starting conditions at the top. If you have used the rhyolite-MELTS graphical user interface (or alphaMELTS) and are used to starting where you just left off then it is pretty easy to accidentally ask MELTS_Excel to perform a huge jump in T or P, which will probably fail. Also note that most calculations work better when run down-temperature (i.e. T1 >= T2).
• If you are doing a calculation that executes a P-T grid, e.g. "Draw Phase Diagram" or "Qtz+Fspars P Calc", then there is a bug that can get MELTS for Excel into an endless loop. Try increasing T1 so that it is above the (wet) liquidus for all pressures. Alternatively, there is a version of MELTS for Excel in the Combine tbl GitList repository that fixes this problem.
• Make sure you are using the correct model, rhyolite-MELTS 1.0.2, 1.1.0 and 1.2.0, or pMELTS. You are given some guidance on the Model tab. Also bear in mind that the solidus and liquidus may be different for the same composition in different models. MELTS for Excel is not geared towards subsolidus calculations.
• Many of the usual tricks for getting MELTS calculations to work apply to the Excel version. For example, you may try varying the T and/or P increments (Delta T and Delta P) if a Sequence seems to be failing.
• The incorporation of some minor components in solid phases in MELTS is oversimplified. Cr2O3, which has a strong effect on spinel stability, is the most notable example. There can also be cases when it is better to set the mass of K2O, MnO, NiO or P2O5 to zero (and consider them as trace elements instead). Note also that, of the oxides mentioned here, only Cr2O3 and K2O are calibrated in the pMELTS model.
• Depending on the composition and conditions, fO2-buffered calculations are sometimes problematic. So try repeating a calculation without the "fO2 Constrained" box checked. If fO2 is not constrained then you need to input the FeO and Fe2O3.
• If you get "Run-time error '13': Type mismatch." and/or the temperature shoots up to > 5000 ^oC then you probably selected "Find Liquidus" for a water-saturated composition. In MELTS 'solids' is used to mean 'not silicate liquid', which includes any water or fluid phase. "Find Wet Liquidus" accounts for this and finds the temperature above which the system is liquid, plus vapour.
• Calculations at relatively low temperatures, involving phases that have ordering (e.g. spinel, pyroxenes, felspathoids) can be slower and there is a possibility that the calculation will time out (see next section).

Server problems

The MELTS for Excel server is located in Seattle, WA, and administered by Mark Ghiorso.

• Sometime a particular thread (MELTS for Excel connection) gets stuck in a loop and becomes unresponsive. If you exit and restart Excel you may pick up another idle thread and find that MELTS for Excel starts working again.
• There is a limit to the number of connections that can be active at one time. During the day, Mark Ghiorso regularly kills and restarts the unresponsive threads. Bear in mind that the server is located on the US west coast - normal service may not resume till the next day for your time zone.
• If the problem persists you could try posting on the Facebook page to see if others are also having (unusual) problems and/or sending a message to Mark Ghiorso (his email is on the MELTS Facebook "About" page), as the server may need attention.
• If MELTS_Excel.log contains "502 Bad Gateway", "503 Service Unavailable" or another HTML error, then this is a server-end problem. More than likely, it is just very busy and if you try again in a few minutes it will be fine.
• If MELTS_Excel.log contains "Time limit reached" then Excel timed out before the server was able to return any results. That is probably an internet connectivity problem, though it may be down to the particular calculation (see previous section). If may also be down to server load, in which case trying again when the server is likely to be less busy may fix the issue.
• When classes are in session, the MELTS for Excel server is likely to be busy during the daytime for US time zones (14:00 GMT - 24:00 GMT, say). To judge from the messages we get, there is also a spike in usage when a large conference is approaching, particularly the AGU Fall Meeting.

Connection problems

Obviously we're not able to offer support for internet connection problems, but one or more of these suggestions may help:

• Slow wireless connections are probably the commonest causes of MELTS_Excel problems. If you can switch to a faster wireless, or wired connection then that should be better.
• Depending on the network, if you can use a connection that does not require a USB or Thunderbolt ethernet adapter (e.g. try a desktop computer) then this may help...
• ... although some university networks have a firewall that interferes with running MELTS_Excel, in which case try it elsewhere.
• If you are able to set up your computer using a static IP, rather than DHCP, then that may make a difference.
• If you know what you're doing, you may consider changing the DNS settings on your computer and/or router.
• If MELTS_Excel usually works, but is timing out on a particular day then simply rebooting your computer and/or router may fix it.

If none of this works, note that MELTS for Excel has a Combine tbl function on the Tools tab that can be used to import text files output by the Rhyolite-MELTS GUI, and automatically create tables and plots similar to those generated by full MELTS for Excel calculations. This "Combine tbl" function is available offline. So for classes or workshops, where connectivity on the day is critical, it may be worth having previously generated output from the GUI available as a backup.

If you experience errors other than "Server returned no results, giving up!" or "Run-time error '13': Type mismatch." do let us know. Also, if you have suggestions of your own for getting MELTS for Excel up and running please do add a comment below.

Thanks!
Paula